PLANTATION YIELD ESTIMATE USING AGRONOMIC NEURAL NETWORK

专利摘要:
in one embodiment, a server computer system receives a set of particular data related to one or more agricultural fields, in which the set of particular data comprises identification data of particular plantations, particular environmental data and data of particular management practices. using a first neural network, the server computer system computes a plantation identification effect on plantation yield from the identification data of particular plantations. using a second neural network, the server computer system computes an environmental effect on plantation yield from particular environmental data. using a third neural network, the server computer system computes a management practice effect on plantation yield from management practice data. using a master neural network, the server computer system computes one or more predicted yield values from the plantation identification effect on plantation yield, the environmental effect on plantation yield and the effect of management practice on plantation yield .
公开号:BR112019015401A2
申请号:R112019015401-9
申请日:2018-01-09
公开日:2020-03-31
发明作者:Guan Wei；Andrejko Erik
申请人:The Climate Corporation；
IPC主号:

专利说明:

PLANTATION YIELD ESTIMATE USING AGRONOMIC NEURAL NETWORK
COPYRIGHT NOTICE [001] A portion of the disclosure in this patent document contains material that is subject to copyright protection. The copyright owner has no objection to facsimile reproduction by anyone of the patent document or patent disclosure as it appears in the patent or trademark file of the Trademark and Patent Office, but otherwise reserves all copyrights or rights whatever they may be. © 2017 The Climate Corporation.
REVELATION FIELD [002] The present disclosure concerns the technical field of computer systems useful in training and executing neural networks. The disclosure also concerns the field of computer systems programmed or configured to use a plurality of neural networks trained with agricultural data as inputs to produce plantation yield values as outputs.
BACKGROUND [003] The approaches described in this section are approaches that can be adopted, but are not necessarily approaches that have been designed or adopted previously. Therefore, unless otherwise stated, it should not be assumed that any of the approaches described in this section are qualified as prior art merely because of their inclusion in this section.
[004] Farmers must often make planting decisions with respect to one or more fields based on
Petition 870190123722, of 11/26/2019, p. 8/105
2/90 incomplete information. In general, the farmer's objective is to maximize plantation yield, quality of planting and / or profits from the sale of the harvest. However, he often has doubts about what combination of crop types, soil types, weather events and management practices will result in maximizing these values.
[005] Agronomic modeling techniques are often used to model interactions between a plantation and the environment. For example, an agronomic model can be used to simulate plantation growth based on the amount of nutrients that plantations receive. Ideally, by using a large number of accurate models, each interaction that affects plantation growth can be simulated, thereby giving perfect knowledge of production results when planting is carried out.
[006] Unfortunately, using such a large number of models to capture every interaction between the plantation and the environment would be computationally expensive. Additionally, the strength of the agronomic model is limited by the knowledge of the person generating the agronomic model. Thus, an agronomic model is unable to consider relationships that are not understood before the model was created.
[007] Neural networks are becoming increasingly popular to solve various types of problems without requiring relationships to be specified in advance. In general, neural networks consist of a series of equations, each of which is configured to transform a plurality of different inputs into one or more outputs. As neural networks are trained, weights
Petition 870190123722, of 11/26/2019, p. 9/105
3/90 are assigned to the series of equations to ensure that the neural network produces correct outputs from the inputs. A benefit of neural networks is that they can capture relationships that are not fully understood by domain experts.
[008] A weakness of neural networks is that they tend to work on a single type of input to produce a single type of output. In the case of agronomic modeling, there are several different types of inputs, including crop type, soil type, effects of weather conditions and management practices, which are relevant to plantation yield. The different types of inputs can be represented differently, as some inputs, such as temperature, vary over time while other inputs, such as soil type, vary spatially.
[009] Thus, there is a need for a comprehensive neural network that can interact with various types of agricultural data in order to provide production results.
SUMMARY [010] The attached claims may serve as a summary of the disclosure.
BRIEF DESCRIPTION OF THE DRAWINGS [011] In the drawings:
[012] Figure 1 illustrates an example computer system that is configured to perform the functions described in this document, shown in a field environment with another device with which the system can operate together.
Petition 870190123722, of 11/26/2019, p. 10/105
4/90 [013] Figure 2 illustrates two views of an example logical arrangement of instruction sets in main memory when an example mobile application is loaded for execution.
[014] Figure 3 illustrates a programmed process by which the agricultural intelligence computer system generates one or more preconfigured agronomic models using agronomic data provided by one or more data sources.
[015] Figure 4 is a block diagram illustrating a computer system in which a modality of the invention can be implemented.
[016] Figure 5 represents an example of a timeline view for data entry.
[017] Figure 6 represents an example modality of a spreadsheet view for data entry.
[018] Figure 7 represents a neural network architecture for computing one or more predicted yield values from one or more plantation-related entries.
[019] Figure 8 represents an example method for executing an agronomic neural network.
DETAILED DESCRIPTION [020] In the following description, for the purpose of explanation, numerous specific details are set out in order to provide a complete understanding of the present disclosure. It will be apparent, however, that modalities can be practiced without these specific details. In other instances, well-known structures and devices are shown in the form of a block diagram in order to
Petition 870190123722, of 11/26/2019, p. 11/105
5/90 avoid unnecessarily obscuring the present revelation. Modalities are revealed in sections according to the following general lines:
1. OVERVIEW
2. AGRICULTURAL INTELLIGENCE COMPUTER SYSTEM OF
EXAMPLE
2.1. STRUCTURAL OVERVIEW
2.2. OVERVIEW OF APPLICATION PROGRAM
2.3. DATA INGESTION FOR THE COMPUTER SYSTEM
2.4. OVERVIEW OF PROCESS - AGRONOMIC MODEL TRAINING
2.5. IMPLEMENTATION EXAMPLE - HARDWARE OVERVIEW
3. AGRONOMIC NEURAL NETWORK
3.1. TRAINING DATA
3.2. NEURAL PLANTATION IDENTIFICATION NETWORK
3.3. NEURAL ENVIRONMENTAL NETWORK
3.3.1. TEMPORAL ENVIRONMENTAL DATA
3.3.2. SPACE ENVIRONMENTAL DATA
3.3.3. DISCOVERED ENVIRONMENTAL INCORPORATION
3.4. NEURAL MANAGEMENT PRACTICE NETWORK
3.5. ADDITIONAL NEURAL NETWORKS
3.6. INTERMEDIATE CROSS-CUT INCORPORATIONS
3.7. MASTER NEURAL NETWORK
3.8. INCOME VALUES
4. APPLICATIONS
4.1. IMPLEMENTATION OF THE NEURAL NETWORK
4.2. RECOMMENDATIONS
Petition 870190123722, of 11/26/2019, p. 10/125
6/90
4.3. PREDICATIONS BASED ON NEW INFORMATION
5. BENEFITS OF CERTAIN MODALITIES
6. EXTENSIONS AND ALTERNATIVES [021] 1. OVERVIEW [022] Systems and methods for generating and using an agronomic neural network configured to use a plurality of different types of data as inputs and producing crop yield values as outputs are described in this document, and provide improved computer-implemented techniques to estimate the yield of agricultural plantations in fields or plantations to be carried out. The agronomic neural network comprises a plurality of individual neural networks configured to produce values indicating an effect on the yield of a plantation. A first neural network is configured to accept plantation identification data as an input and generate a plantation identification effect at full yield as an output. A second neural network is configured to accept environmental data as input and to compute an environmental effect in total output as output. A third neural network is configured to accept data from management practices as input and compute an effect of management practice at full throughput as output. A master neural network is configured to accept the plantation identification effect at full yield, the environmental effect at full yield and the effect of management practice at full yield as inputs and produce one or more yield values as outputs.
[023] In one embodiment, a method comprises receiving, in a server computing system, a set of
Petition 870190123722, of 11/26/2019, p. 10/13
7/90 private data relating to one or more agricultural fields, where the private data set comprises identification data for particular plantations, private environmental data and private management practice data; using a first neural network configured using plantation identification data as input and plantation yield data as output, compute a plantation identification effect on plantation yield for the one or more agricultural fields from the identification data of particular plantations ; using a second neural network configured using environmental data as input and plantation yield data as an output, to compute an environmental effect on plantation yield for the one or more agricultural fields from the particular environmental data; using a third neural network configured using management practice data as input and plantation yield data as output, to compute a management practice effect on plantation yield for the one or more agricultural fields from the data of particular management practices ; using a master neural network configured using plantation identification effects on plantation yield, environmental effects on plantation yield and management practice effects on plantation yield as input and plantation yield data as output, compute one or more values of yields predicted for one or more agricultural fields from the plantation identification effect to plantation yield, the environmental effect on plantation yield and the
Petition 870190123722, of 11/26/2019, p. 10/145
8/90 crop yield management practice.
[024] 2. EXAMPLE AGRICULTURAL INTELLIGENCE COMPUTER SYSTEM [025] 2.1 STRUCTURAL OVERVIEW [026] Figure 1 illustrates an example computer system that is configured to perform the functions described in this document, shown in a field environment with another device with which the system can operate together. In one embodiment, a user 102 owns, operates, or controls a field manager computing device 104 at a field location or associated with a field location such as an intended field for agricultural activities or a management location for one or more fields agricultural Field manager computing device 104 is programmed or configured to provide field data 106 to an agricultural intelligence computer system 130 via one or more networks 109.
[027] Examples of field data 106 include (a) identification data (for example, area size in acres, field name, field identifiers, geographical identifiers, boundary identifiers, plantation identifiers and any other appropriate data that can be used to identify farmland, such as a common land unit (CLU), lot and block number, a land number, geographic coordinates and boundaries, Farm Serial Number (FSN), farm number, number section number, field number, section, municipal district and / or strip), (b) harvest data (for example, crop type, crop variety, crop rotation, if the
Petition 870190123722, of 11/26/2019, p. 10/15
9/90 plantation is organically grown, harvest date, Current Production History (APH), expected yield, yield, crop price, harvest recipe, grain moisture, cultivation practice and previous cultivation season information, (c ) soil data (eg type, composition, pH, organic matter (OM), cation exchange capacity (CEC)), (d) planting data (eg planting date, seed type (s) (s), relative maturity (RM) of planted seed (s), seed population), (e) fertilizer data (eg nutrient type (Nitrogen, Phosphorus, Potassium), type of application, date of application, quantity, source, method), (f) pesticide data (for example, pesticide, herbicide, fungicide, other substance or mixture of substances intended for use as a plant regulator, defoliant or desiccant, date of application, quantity, source , method), (g) irrigation data (for example, application date, quantity, source, method), (h) meteor data (eg rainfall, rainfall index, forecast rain, water runoff region, temperature, wind, forecast, pressure, visibility, clouds, heat index, dew point, humidity, snow depth, quality of air, sunrise, sunset), (i) image data (for example, images and light spectrum information from a farm appliance sensor, camera, computer, smart phone, tablet, unmanned aerial vehicle, planes or satellite), (j) exploration observations (photos, videos, free-form notes, voice recordings, voice transcriptions, weather conditions (temperature, precipitation (current and over time), soil moisture,
Petition 870190123722, of 11/26/2019, p. 10/165
10/90 plantation growth, wind speed, relative humidity, dew point, black layer)), and (k) soil, seed, crop phenology, pest and disease reports and sources of predictions and databases.
[028] A data server computer 108 is communicatively coupled to the agricultural intelligence computer system 130 and is programmed or configured to send external data 110 to the agricultural intelligence computer system 130 via the network (s) 109. External data server computer 108 may be owned or operated by the same person or legal entity such as the agricultural intelligence computer system 130, or by a different person or entity such as a government agency, non-governmental organization (NGO ) and / or a private data service provider. Examples of external data include meteorological data, image data, soil data or statistical data relating to plantation yields, among others. External data 110 may consist of the same type of information as field data 106. In some embodiments, external data 110 is provided by an external data server 108 belonging to the same entity that owns and / or operates the intelligence computer system. agricultural 130. For example, the agricultural intelligence computer system 130 may include a data server focused exclusively on one type of data that may otherwise be obtained from sources of external entities, such as meteorological data. In some embodiments, an external data server 108 can actually be incorporated into system 130.
Petition 870190123722, of 11/26/2019, p. 10/175
11/90 [029] An agricultural device 111 can have one or more remote sensors 112 incorporated in it, whose sensors are communicatively coupled directly or indirectly via agricultural device 111 to the agricultural intelligence computer system 130 and are programmed or configured to send sensor data to the agricultural intelligence computer system 130. Examples of farm equipment 111 include tractors, combine harvesters, harvesters, seeders, trucks, fertilizer equipment, unmanned aerial vehicles, and any other item of physical machinery or hardware, typically mobile machinery, and that can be used in tasks associated with agriculture. In some embodiments, a single unit of apparatus 111 may comprise a plurality of sensors 112 that are coupled locally to a network in the apparatus; Area control network (CAN) is an example of a network like this that can be installed on combine harvesters or harvesters. Application controller 114 is communicatively coupled to the agricultural intelligence computer system 130 via network (s) 109 and is programmed or configured to receive one or more scripts to control an operating parameter of a vehicle or agricultural implement of the agricultural intelligence computer system 130. For example, a network bus interface of
Control of area (CAN) can be used to empower communications From istema in computer in intelligence agricultural 130 for The device agricultural 111, like the
CLIMATE FIELDVIEW DRIVE, available from The Climate
Corporation, San Francisco, California, is used. Data from
Petition 870190123722, of 11/26/2019, p. 10/185
12/90 sensors may consist of the same type of information as field data 106. In some embodiments, remote sensors 112 may not be incorporated into an agricultural appliance 111 and may be located remotely in the field and may communicate with network 109.
[030] Apparatus 111 may comprise a cabin computer 115 that is programmed with a cabin application, which may comprise a version or variant of the mobile application for device 104 which is further described in other sections in this document. In one embodiment, the cabin computer 115 comprises a compact computer, often a computer the size of a tablet or smart phone, with a graphical screen display, such as a color display, which is mounted within the operator cabin of the apparatus 111. Cabin computer 115 can implement all or some of the operations and functions that are further described in this document for mobile computing device 104.
[031] Network (s) 109 generally represents any combination of one or more data communication networks including local area networks, wide area networks, network interconnections or internets, using any one of wired or wireless links, including terrestrial or satellite links. The network (s) can be implemented by any medium or mechanism that allows the exchange of data between the various elements of figure 1. The various elements of figure 1 can also have direct communication links ( wired or wireless). Each of the sensors 112, the controller 114, the external data server computer 108 and other elements of the system comprises an interface
Petition 870190123722, of 11/26/2019, p. 10/195
13/90 compatible with network (s) 109 and is programmed or configured to use standardized protocols for communication over networks such as TCP / IP, Bluetooth, CAN protocol and higher layer protocols such as HTTP, TLS and others more.
[032] The agricultural intelligence computer system 130 is programmed or configured to receive field data 106 from field manager computing device 104, external data 110 from external data server computer 108 and remote sensor sensor data 112 The agricultural intelligence computer system 130 can be further configured to host, use or run one or more computer programs, other software elements, digitally programmed logic such as FPGAs or ASICs, or any combination of these to perform translation and storage of data values, building digital models of one or more plantations in one or more fields, generating recommendations and notifications, and generating and sending scripts to application controller 114, in the manner described further in other sections of this disclosure.
[033] In one embodiment, the agricultural intelligence computer system 130 is programmed with or comprises a communication layer 132, the presentation layer 134, the data management layer 140, the hardware / virtualization layer 150 and the repository of model and field data 160. Layer, in this context, refers to any combination of digital electronic interface circuits, microcontrollers, firmware such as drivers and / or computer programs or other
Petition 870190123722, of 11/26/2019, p. 10/20
14/90 software elements.
[034] Communication layer 132 can be programmed or configured to perform connection functions via the input / output interface including sending requests to the field manager computing device 104, the external data server computer 108 and the sensor remote 112 for field data, external data and sensor data respectively. Communication layer 132 can be programmed or configured to send received data to the model and field data repository 160 to be stored as field data 106.
[035] Presentation layer 134 can be programmed or configured to generate a graphical user interface (GUI) to be displayed on the field manager computing device 104, on the cabin computer 115 or on other computers that are coupled to the system 130 through network 109. The GUI can comprise controls for entering data to be sent to the agricultural intelligence computer system 130, generating requests for models and / or recommendations and / or displaying recommendations, notifications, models and other field data.
[036] The data management layer 140 can be programmed or configured to manage read operations and write operations involving the repository 160 and other functional elements of the system, including queries and result sets transmitted between the functional elements of the system and the repository. Examples of data management layer 140 include JDBC, SQL server interface code and / or code
Petition 870190123722, of 11/26/2019, p. 10/21
15/90 HADOOP interface, among others. The repository 160 can comprise a database. As used in this document, the term database can refer to a dataset, a relational database management system (RDBMS) or both. As used in this document, a database can comprise any collection of data including hierarchical databases, relational databases, flat file databases, object relational databases, object oriented databases and any other structured collection of records or data that is stored on a computer system. Examples of RDBMSs include, but are not limited to, ORACLE®, MYSQL, IBM® DB2, MICROSOFT® SQL SERVER, SYBASE® and POSTGRESQL databases. However, any database that enables the systems and methods described in this document can be used.
[037] When field data 106 is not provided directly to the agricultural intelligence computer system through one or more agricultural machines or agricultural machinery devices that interact with the agricultural intelligence computer system, the user can be guided through one or more user interfaces on the user device (served by the agricultural intelligence computer system) to enter such information. In an example mode, the user can specify identification data when accessing a map on the user's device (served by the agricultural intelligence computer system) and select specific CLUs that have been shown graphically on the map. In an alternative mode, user 102 can specify
Petition 870190123722, of 11/26/2019, p. 10/22
16/90 identification data when accessing a map on the user's device (served by the agricultural intelligence computer system 130) and plotting field boundaries on the map. Such a selection of CLU or map drawings represent geographical identifiers. In alternative modalities, the user can specify identification data when accessing field identification data (provided as shape files or in a similar format) from the US Department of Agriculture's Rural Services Agency or another source through the tracking device. user and provide such field identification data to the agricultural intelligence computer system.
[038] In an example embodiment, the agricultural intelligence computer system 130 is programmed to generate and display a graphical user interface comprising a data manager for data entry. After one or more fields have been identified using the methods described above, the data manager can provide one or more graphical user interface symbols that when selected can identify changes to the field, soil, crops, crops or nutrient practices. The data manager can include a timeline view, a spreadsheet view and / or one or more editable programs.
[039] Figure 5 represents an example modality of a timeline view for data entry. Using the display shown in figure 5, a user computer can enter a selection of a particular field and a particular date for adding an event. Events represented at the top of the timeline
Petition 870190123722, of 11/26/2019, p. 10/23
17/90 may include Nitrogen, Planting, Practices and Soil. To add a nitrogen application event, a user computer can provide input to select the nitrogen tab. The user computer can then select a location on the timeline for a particular field to indicate an application of nitrogen to the selected field. In response to receiving a selection from a location on the timeline for a particular field, the data manager may display a data entry overlay, allowing the user's computer to enter data relating to nitrogen applications, planting procedures, application soil, tillage procedures, irrigation practices or other information relating to the particular field. For example, if a user's computer selects a portion of the timeline and indicates a nitrogen application, then the data entry overlay may include fields to enter an amount of applied nitrogen, an application date, a type of fertilizer used and any other information related to the application of nitrogen.
[040] In one embodiment, the data manager provides an interface for creating one or more programs. Program, in this context, refers to a set of data related to nitrogen applications, planting procedures, application in the soil, tillage procedures, irrigation practices or other information that may be related to one or more fields, and that can be stored in digital data storage for reuse as a set in other operations. After a program has been created, it can be applied conceptually to one or
Petition 870190123722, of 11/26/2019, p. 10/24
18/90 more fields and references to the program can be stored in digital storage in association with data identifying the fields. So, instead of manually inputting identical data relating to the same nitrogen applications for multiple different fields, a user computer can create a program that indicates a particular application of nitrogen and then can apply the program to multiple different fields. For example, in the timeline view in Figure 5, the top two timelines have the Applied Autumn program selected, which includes an application of 150 1b (68.08 kg) N / ac in early April. The data manager can provide an interface for editing a program. In one mode, when a particular program is edited, each field that selected the particular program is edited. For example, in figure 5, if the Autumn Applied program is edited to reduce nitrogen application to 130 1b (58.97 kg) N / ac, the top two fields can be
updated with an application reduced from nitrogen with base in the program edited. [041] In a modality, in response to receive edits for One field that has a program selected, the
Data manager removes the field match for the selected program. For example, if a nitrogen application is added to the upper field in figure 5, the interface may update to indicate that the Autumn Applied program is no longer being applied to the upper field. Although the application of nitrogen in early April may remain, updates to the Applied program in the fall would not change the application of
Petition 870190123722, of 11/26/2019, p. 10/25
19/90 nitrogen in April.
[042] Figure 6 represents an example of a spreadsheet view for data entry. Using the display shown in figure 6, a user can create and edit information for one or more fields. The data manager can include spreadsheets to enter information regarding Nitrogen, Planting, Practices and Soil as shown in figure 6. To edit a particular entry, a user computer can select the particular entry in the spreadsheet and update the values. For example, figure 6 represents an ongoing update to a target yield value for the second field. Additionally, a user's computer can select one or more fields in order to apply one or more programs. In response to receiving a program selection for a particular field, the data manager can automatically complete entries for the particular field based on the selected program. As with the timeline view, the data manager can update entries for each field associated with a particular program in response to receiving an update for the program. In addition, the data manager can de-match the program selected for the field in response to receiving an edit for one of the entries for the field.
[043] In one embodiment, model and field data is stored in the model and field data repository 160. Model data comprises data models created for one or more fields. For example, a plantation model may include a digitally constructed model from
Petition 870190123722, of 11/26/2019, p. 10/26
20/90 development of a plantation in one or more fields. Model, in this context, refers to a digitally electronic stored set of executable instructions and data values, associated with each other, which are capable of receiving and responding to an invocation, request or programmatic or other digital call for resolution based at specified input values, to produce one or more stored output values that can serve as the basis for computer-implemented recommendations, displays of output data or machine control, among other things. People with professional knowledge in the field know that it is convenient to express models using mathematical equations, but this form of expression does not limit the models revealed in this document to abstract concepts; instead, each model in this document has a practical application on a computer in the form of executable instructions and stored data that implement the model using the computer. The model can include a model of past events in one or more fields, a model of the current status of one or more fields and / or a model of events predicted in one or more fields. Model and field data can be stored in memory data structures, rows in a database table, in simple files or spreadsheets, or in other forms of stored digital data.
[044] The neural network training instructions 136 comprise one or more instructions that, when executed by the agricultural intelligence computer system 130, induce the agricultural intelligence computer system 130 to train a neural network using a plurality of
Petition 870190123722, of 11/26/2019, p. 10/27
21/90 data sets, each comprising plantation identification data, environmental data, management practice data and one or more yield values. The neural network execution instructions 138 comprise one or more instructions that, when executed by the agricultural intelligence computer system 130, induce the agricultural intelligence computer system 130 to use the trained neural network to compute one or more yield values of plantations from a particular data set comprising identification data from particular plantations, particular environmental data and data from particular management practices.
[045] In one embodiment, each of the neural network training instructions 136 and the neural network execution instructions 138 comprises a set of one or more pages of main memory, such as RAM, in the agricultural intelligence computer system. 130 in which executable instructions have been loaded and which, when executed, induce the agricultural intelligence computing system to perform the functions or operations that are described in this document with reference to those modules. For example, neural network training instructions 136 may comprise a set of RAM pages that contain instructions that, when executed, induce execution of the neural network training functions that are described in this document. The instructions may be in machine executable code in a CPU's instruction set and may have been compiled based on source code written in JAVA, C, C ++, OBJECTIVE-C, or any
Petition 870190123722, of 11/26/2019, p. 10/285
22/90 another human-readable programming language or environment, alone or in combination with JAVASCRIPT scripts, other written languages and other programming source text. The term pages is proposed to refer in general to any region within main memory and the specific terminology used in a system may vary depending on the memory architecture or processor architecture. In another embodiment, each of the neural network training instructions 136 and the neural network execution instructions 138 may also represent one or more files or source code projects that are stored digitally on a mass storage device such as Non-volatile RAM or disk storage, in the agricultural intelligence computer system 130 or in a separate repository system, which when compiled or interpreted induce the generation of executable instructions that, when executed, induce the agricultural intelligence computing system to perform the functions or operations that are described in this document with reference to those modules. In other words, the figure can represent the way in which programmers or software developers organize and arrange source code for later compilation into an executable, or interpretation in byte code or the equivalent, for execution by the agricultural intelligence computer system 130 .
[046] The hardware / virtualization layer 150 comprises one or more central processing units (CPUs), memory controllers and other devices, components or elements of a computer system, such as memory
Petition 870190123722, of 11/26/2019, p. 10/29
23/90 volatile or non-volatile, non-volatile storage such as disk, and input / output devices or interfaces as illustrated and described, for example, in connection with figure 4. Layer 150 can also comprise programmed instructions that are configured to support virtualization, containerization or other technologies.
[047] For the purpose of illustrating a clear example, figure 1 shows a limited number of instances of certain functional elements. However, in other modalities, there may be any number of such elements. For example, modalities can use thousands or millions of different mobile computing devices 104 associated with different users. Additionally, system 130 and / or the external data server computer 108 can be implemented using two or more processors, cores, clusters or instances of physical machines or virtual machines, configured in a different location or located together with other elements in a center data, shared computing structure or cloud computing structure.
[048] 2.2. OVERVIEW OF APPLICATION PROGRAM [049] In one embodiment, the implementation of the functions described in this document using one or more computer programs or other software elements that are loaded and executed using one or more general purpose computers will induce the computers in use general to be configured as a particular machine or as a computer that is specially adapted to perform the functions described in this document. In addition, each of the flow diagrams that are described further
Petition 870190123722, of 11/26/2019, p. 10/30
24/90 in this document can serve, alone or in combination with the descriptions of processes and functions exposed in this document, as algorithms, plans or directions that can be used to program a computer or logic to implement the functions that are described. In other words, all of the text exposed in this document and all figures are presented together to provide revelation of algorithms, plans or directions that are sufficient to allow a qualified person to program a computer to perform the functions that are described in this document, in combination with the professional and practical knowledge of a person like this considering the level of professional knowledge that is appropriate for inventions and revelations of this type.
[050] In one embodiment, user 102 interacts with the agricultural intelligence computer system 130 using the field manager computing device 104 configured with an operating system and one or more application programs or applications; the field manager computing device 104 can also operate in conjunction with the agricultural intelligence computer system independently and automatically under program control or logical control and direct user interaction is not always required. The field manager computing device 104 generally represents one or more of a smart phone, PDA, tablet, laptop, desktop computer, workstation, or any other computing device capable of transmitting and receiving information and performing functions described in this document. The field manager computing device 104 can
Petition 870190123722, of 11/26/2019, p. 10/31
25/90 communicating over a network using a mobile application stored in the field manager computing device 104, and in some embodiments the device can be coupled to sensor 112 and / or controller 114 using a cable 113 or connector. A private user 102 may own, operate or control and use, in connection with system 130, more than one field manager computing device 104 at a time.
[051] The mobile application can provide client-side functionality, over the network, for one or more mobile computing devices. In an example embodiment, the field manager computing device 104 can access the mobile application via a web browser or a local client application or application. Field manager computing device 104 can transmit data to, and receive data from, one or more client-side servers, using network-based protocols or formats such as HTTP, XML and / or JSON, or application-specific protocols. In an example embodiment, the data may take the form of requests and input of user information, such as field data, to the mobile computing device. In some embodiments, the mobile application interacts with location tracking hardware and software on the field manager computing device 104, which determines the location of the field manager computing device 104 using standard tracking techniques such as multilateration of radio signals , the global positioning system (GPS), WiFi positioning systems, or other mobile positioning methods. In some
Petition 870190123722, of 11/26/2019, p. 10/32
26/90 cases, location data or other data associated with device 104, user 102 and / or user account (s) can be obtained by consulting a device's operating system or by requesting an application on the device to get data from the operating system.
[052] In one embodiment, the field manager computing device 104 sends field data 106 to the agricultural intelligence computer system 130 comprising or including, but not limited to, data values representing one or more of: a geographical location of one or more fields, crop information for one or more fields, plantations carried out in one or more fields and soil data extracted from one or more fields. Field manager computing device 104 can send field data 106 in response to user input from user 102 by specifying data values for the one or more fields. In addition, field manager computing device 104 can automatically send field data 106 when one or more of the data values become available to field manager computing device 104. For example, field manager computing device 104 can be communicatively coupled to remote sensor 112 and / or application controller 114. In response to receiving data indicating that application controller 114 has released water for one or more fields, the field manager computing device 104 can send data field 106 for the agricultural intelligence computer system 130 indicating that water has been released to the one or more fields. The field data 106 identified in this disclosure can
Petition 870190123722, of 11/26/2019, p. 10/33
27/90 be introduced and transmitted using electronic digital data that is transmitted between computing devices using URLs parameterized in HTTP, or another suitable communication or message protocol.
[053] A commercial example of the mobile application is CLIMATE FIELDVIEW, commercially available from The Climate Corporation, San Francisco, California. The CLIMATE FIELDVIEW application, or other applications, may be modified, extended or adapted to include features, functions and programming that were not revealed prior to the filing date of this disclosure. In one embodiment, the mobile application comprises an integrated software platform that allows a farmer to make fact-based decisions for his operation because it combines historical data about the farmer's fields with any other data with which the farmer wants to compare. Combinations and comparisons can be performed in real time and are based on scientific models that provide potential scenarios to enable the farmer to make better decisions while being better informed.
[054] Figure 2 illustrates two views of an example logical arrangement of instruction sets in main memory when an example mobile application is loaded for execution. In figure 2, each named element represents a region of one or more pages of RAM or other main memory, or one or more blocks of disk storage or other non-volatile storage, and the instructions programmed within those regions. In one embodiment, in view (a), a mobile computer application 200 comprises the account instructions
Petition 870190123722, of 11/26/2019, p. 10/34
28/90 fields-data ingestion-sharing 202, overview and alert instructions 204, digital map book instructions 206, seed and planting instructions 208, nitrogen instructions 210, weather instructions 212 , field health instructions 214 and performance instructions 216.
[055] In one embodiment, a mobile computer application 200 comprises account-field-data-ingestion-sharing instructions 202 that are programmed to receive, translate and ingest data from external entity systems' fields via manual transfer or APIs. Types of data may include field boundaries, production maps, maps such as planted, soil test results, maps such as applied and / or management zones, among others. Data formats can include shape files, natural data formats from external entities and / or exports from the farm management information system (FMIS), among others. Data receipt can occur via manual transfer, email with attachment, external APIs that push data to the mobile application, or instructions that call APIs from external systems to pull data into the mobile application. In one embodiment, the mobile computer application 200 comprises a data entry box. In response to receiving a selection from the data entry box, the mobile computer application 200 may display a graphical user interface for manually transferring data files and importing transferred files into a data manager.
[056] In one embodiment, the book instructions for
Petition 870190123722, of 11/26/2019, p. 10/35
29/90 digital maps 206 comprise data layers of field maps stored in device memory and are programmed with data visualization tools and geospatial field notes. This provides farmers with convenient information very easy for reference, login and visual understandings for field performance. In one embodiment, the overview and alert instructions 204 are programmed to provide a broad operating view of what is important to the farmer, and recommendations at the right time to take action or focus on particular problems. This allows the farmer to focus in time on what needs attention, to save time and preserve production throughout the season. In one embodiment, seed and planting instructions 208 are programmed to provide tools for seed selection, placement of hybrids, and creation of scripts, including creation of variable rate (VR) scripts, based on scientific models and empirical data. This enables farmers to maximize production or return on investment through the purchase, placement and optimized seed population.
[057] In one embodiment, script generation instructions 205 are programmed to provide an interface for generating scripts, including variable rate (VR) fertility scripts. The interface enables farmers to create scripts for field implements, such as nutrient applications, planting and irrigation. For example, a planting script interface can comprise tools to identify a type of seed to plant. Upon receiving a selection of the seed type, the
Petition 870190123722, of 11/26/2019, p. 36/105
30/90 mobile computer application 200 can display one or more fields divided into management zones, such as the layers of field map data created as part of the digital map book instructions 206. In one embodiment, the management zones comprise soil zones together with a panel identifying each soil zone and a soil name, texture, drainage for each zone or other field data. The mobile computer application 200 can also display tools for editing or creating, such as graphical tools for drawing management zones, such as soil zones, on a map of one or more fields. Planting procedures can be applied to all management zones or different planting procedures can be applied to different subsets of management zones. When a script is created, the mobile computer application 200 can make the script available for download in a format readable by an application controller, such as an archived or compressed format. Additionally, and / or alternatively, a script can be sent directly to the cabin computer 115 by the mobile computer application 200 and / or transferred to one or more data servers and stored for further use.
[058] In one embodiment, nitrogen instructions 210 are programmed to provide tools to inform about nitrogen decisions when viewing nitrogen availability for plantations. This enables farmers to maximize production or return on investment through optimized nitrogen application during the season. Sample programmed functions include
Petition 870190123722, of 11/26/2019, p. 37/105
31/90 display images such as SSURGO images to enable design of application zones and / or images generated from subfield soil data, such as data obtained from sensors, in a high spatial resolution (as refined as 10 meters or less because of proximity to the ground); transfer of existing farmer-defined areas; provide an application graph and / or a map to enable adjustment of nitrogen application (s) across multiple zones; production of scripts to drive machinery; tools for entering and adjusting mass data; and / or maps for data visualization, among others. Mass data entry, in this context, may mean entering data once and then applying the same data to multiple fields that have been defined in the system; Sample data may include nitrogen application data that is the same for many fields from the same farmer, but such mass data entry applies to the entry of any type of field data in the mobile computer application 200. For example, nitrogen instructions 210 can be programmed to accept definitions of application programs and nitrogen practices and to accept user input specifying application of these programs across multiple fields. Nitrogen application programs, in this context, refer to a set of named and stored data that associate: a name, color code or other identifier, one or more application dates, types of material or product for each of the dates and quantities, method of application or incorporation as injected or excavated and / or quantities or rates of application
Petition 870190123722, of 11/26/2019, p. 38/105
32/90 for each of the dates, plantation or hybrid that the application aims at, among others. Nitrogen practice programs, in this context, refer to a set of named and stored data that associate: a practice name; a previous plantation; a tillage system; a farming date primarily; one or more previous crop systems that were used; one or more application type indicators, such as fertilizer, that was used. Nitrogen instructions 210 can also be programmed to generate and induce display of a nitrogen graph, which indicates projections of use per plant of the specified nitrogen and whether a surplus or supplement is predicted; in some modalities, different colored indicators may signal a magnitude of surplus or magnitude of complement. In one embodiment, a nitrogen graph comprises a graphical display on a computer display device comprising a plurality of lines, each line associated with and identifying a field; data specifying what is planted in the field, the size of the field, the location of the field and a graphical representation of the field perimeter; in each line, one time line per month with graphic indicators specifying each application and amount of nitrogen at points correlated with month names; and numerical and / or colored indicators of surplus or complement, in which color indicates magnitude.
[059] In one embodiment, the nitrogen graph can include one or more user input features, such as dials or slider bars, to dynamically change application programs and application practices.
Petition 870190123722, of 11/26/2019, p. 39/105
33/90 nitrogen so that a user can optimize their nitrogen graph. The user can then use their optimized nitrogen graph and related nitrogen application and practice programs to implement one or more scripts, including variable rate (VR) fertility scripts. Nitrogen instructions 210 can also be programmed to generate and induce display of a nitrogen map, which indicates projections of use per plant of the specified nitrogen and whether a surplus or supplement is predicted; in some modalities, different colored indicators may signal a magnitude of surplus or magnitude of complement. The nitrogen map can display projections of use per plant of the specified nitrogen and whether a surplus or complement is predicted for different times in the past and in the future (such as daily, weekly, monthly or yearly) using numerical and / or colored surplus indicators or complement, in which color indicates magnitude. In one embodiment, the nitrogen map can include one or more user input features, such as dials or slider bars, to dynamically change nitrogen application and practice programs so that a user can optimize their nitrogen map. nitrogen, such as to obtain a preferred amount of surplus or complement. The user can then use their optimized nitrogen map and related nitrogen application and practice programs to implement one or more scripts, including variable rate (VR) fertility scripts. In other embodiments, instructions similar to nitrogen 210 instructions can be used to
Petition 870190123722, of 11/26/2019, p. 40/105
34/90 application of other nutrients (such as phosphorus and potassium), application of pesticide and irrigation programs.
[060] In one embodiment, weather instructions 212 are programmed to provide recent field-specific weather data and predicted weather information. This enables farmers to save time and have an efficient integrated view of daily operational decisions.
[061] In one embodiment, field health instructions 214 are programmed to provide remote sensing images at the right time highlighting plantation variation at the station and potential concerns. Programmed example functions include cloud checking to identify possible clouds or cloud shading; determination of nitrogen indices based on field images; graphical visualization of exploration layers, including, for example, those related to field health, and observation and / or sharing of exploration notes; and / or transferring satellite images from multiple sources and prioritizing the images for the farmer, among others.
[062] In one embodiment, performance instructions 216 are programmed to provide reports, analyzes and understanding tools using data about the farm for assessment, understandings and decisions. This enables the farmer to look for improved results for the coming year through fact-based conclusions as to why return on investment stayed at previous levels, and understanding for limiting factors of production. Performance instructions
Petition 870190123722, of 11/26/2019, p. 41/105
35/90
216 can be programmed to communicate over network (s) 109 with server-side analytical programs run on the agricultural intelligence computer system 130 and / or on the external data server computer 108 and configured to analyze metrics such as production, hybrid, population, SSURGO, soil tests, or elevation, among others. Scheduled reports and analyzes may include analysis of production variability, production benchmarking and other metrics against other farmers based on collected data kept anonymous from many farmers, or data for seeds and planting, among others.
[063] Applications having instructions configured in this way can be implemented for different computing device platforms while maintaining the same external form of the common user interface. For example, the mobile application can be programmed to run on tablets, smart phones or server computers that are accessed using browsers on client computers. In addition, the mobile application as configured for tablets or smart phones can provide a complete application experience or a cabin application experience that is suitable for the display and processing capabilities of the cabin computer 115. For example, referring now in view (b) of figure 2, in one embodiment a cabin computer application 220 can comprise cabin map instructions 222, remote view instructions 224, data collection and transfer instructions 226, alert instructions machine 228, the transfer instructions for
Petition 870190123722, of 11/26/2019, p. 42/105
36/90 script 230 and the scan-cabin instructions 232. The code base for the instructions in the view (b) can be the same as in the view (a) and executables implementing the code can be programmed to detect the type of platform on the which they are running and to display, through a graphical user interface, only those functions that are appropriate for a cabin platform or complete platform. This approach empowers the system to distinctly recognize the different user experience that is appropriate for a cabin environment and the different cabin technology environment. The mapascabine 222 instructions can be programmed to provide views of maps of fields, farms or regions that are useful when directing machine operation. Remote view instructions 224 can be programmed to connect, manage and provide machine activity views in real time or near real time to other computing devices connected to system 130 via wireless networks, wired connectors or adapters and others more. Data collection and transfer instructions 226 can be programmed to connect, manage and provide transfer of data collected from sensors and controllers to system 130 via wireless networks, wired connectors or adapters and more. Machine alert instructions 228 can be programmed to detect issues with machine or tool operations that are associated with the cab and generate operator alerts. Script transfer instructions 230 can be configured to transfer instruction scripts that are configured to direct machine operations or to
Petition 870190123722, of 11/26/2019, p. 43/105
37/90 data collection. Exploration-cabin instructions 230 can be programmed to display alerts based on locations and information received from system 130 based on the location of agricultural device 111 or sensors 112 in the field and ingest, manage and provide exploration observation transfers based on locations for system 130 based on the location of agricultural device 111 or sensors 112 in the field.
[064] 2.3. DATA INGESTION FOR THE COMPUTER SYSTEM [065] In one embodiment, the external data server computer 108 stores external data 110, including soil data representing soil composition for the one or more fields and meteorological data representing temperature and precipitation in the one or more fields. Weather data can include past and present weather data as well as forecasts for future weather data. In one embodiment, the external data server computer 108 comprises a plurality of servers hosted by different entities. For example, a first server may contain data on soil composition while a second server may include weather data. In addition, soil composition data can be stored on multiple servers. For example, one server can store data representing percentages of sand, silt and clay in the soil while a second server can store data representing percentages of organic matter (OM) in the soil.
[066] In one embodiment, remote sensor 112 comprises one or more sensors that are programmed or configured
Petition 870190123722, of 11/26/2019, p. 44/105
38/90 to produce one or more observations. Remote sensor 112 can be aerial sensors, such as satellites, vehicle sensors, planting equipment sensors, crop sensors, fertilizer or insecticide application sensors, harvester sensors and any other implement capable of receiving data from one or more fields. In one embodiment, application controller 114 is programmed or configured to receive instructions from the agricultural intelligence computer system 130. Application controller 114 can also be programmed or configured to control an operating parameter of a vehicle or agricultural implement. For example, an application controller can be programmed or configured to control a vehicle's operating parameter, such as a tractor, planting equipment, tillage equipment, fertilizer or insecticide equipment, harvesting equipment or other agricultural implements such as like a water valve. Other modalities can use any combination of sensors and controllers, of which the following are merely selected examples.
[067] System 130 can obtain or ingest data, under user control 102, from a mass base of a large number of farmers who have contributed data to a shared database system. This way of obtaining data can be called manual data ingestion since one or more user-controlled computer operations are requested or activated to obtain data for use by the 130 system. As an example, the CLIMATE FIELDVIEW application, commercially available from The Climate Corporation, San Francisco, California, can
Petition 870190123722, of 11/26/2019, p. 45/105
39/90 be operated to export data to system 130 for storage in repository 160.
[068] For example, seed monitoring systems can either control sowing device components or obtain planting data, including signals from seed sensors via a bundle of signal wires comprising a CAN core network and point-to-point connections for registration and / or diagnostics. Seed monitoring systems can be programmed or configured to display seed spacing, population and other information to the user via the cabin computer 115 or other devices within the 130 system. Examples are disclosed in US patent 8,738,243 and in US patent publication 20150094916, and the present disclosure acknowledges these other patent disclosures.
[069] Similarly, production monitoring systems may contain production sensors for harvesting devices that send production measurement data to the cabin computer 115 or to other devices within the 130 system. Production monitoring systems may use one or more more remote sensors 112 to obtain grain moisture measurements on a combine harvester or other harvester and can transmit these measurements to the user via the cabin computer 115 or other devices within the system 130.
[070] In one embodiment, examples of sensors 112 that can be used with any moving vehicle or device of the type described elsewhere in this document include kinematic sensors and position sensors.
Petition 870190123722, of 11/26/2019, p. 46/105
40/90
Kinematic sensors can comprise any of the speed sensors such as radar or wheel speed sensors, accelerometers, or gyroscopes. Position sensors can comprise GPS receivers or transceivers, or WiFi-based positioning or mapping applications that are programmed to determine location based on nearby WiFi access points, among others.
[071] In one embodiment, examples of 112 sensors that can be used with tractors or other moving vehicles include engine speed sensors, fuel consumption sensors, area meters or distance meters that interact with GPS or radar signals. , speed PTO (PTO) sensors, tractor hydraulic sensors configured to detect hydraulic parameters such as pressure or flow, and / or hydraulic pump speed, wheel speed sensors or wheel slip sensors. In one embodiment, examples of controllers 114 that can be used with tractors include hydraulic directional controllers, pressure controllers and / or flow controllers; speed controllers for hydraulic pumps; speed controllers or regulators; obstacle position controllers; or wheel position controllers to provide automatic steering.
[072] In one embodiment, examples of sensors 112 that can be used with seed planting equipment such as seeders, drills, or air seeders include seed sensors, which can be optical, electromagnetic or impact sensors; downward force sensors such as load pins, load cells,
Petition 870190123722, of 11/26/2019, p. 47/105
41/90 pressure sensors; sensors for soil properties such as reflectivity sensors, humidity sensors, electrical conductivity sensors, optical residue sensors or temperature sensors; component operating criteria sensors such as depth sensors
planting, pressure sensors cylinder in strength to low sensors of speeds of discs in seeds, coders engines drive in seeds,
speed sensors of seed conveyor systems, or vacuum level sensors; or sensors for pesticide applications such as optical sensors or other electromagnetic sensors, or impact sensors. In one embodiment, examples of controllers 114 that can be used with such seed planting equipment include: toolbar folding controllers, such as controllers for valves associated with hydraulic cylinders; downward force controllers, such as controllers for valves associated with pneumatic cylinders, airbags or hydraulic cylinders, and programmed to apply downward force to individual row units or to a full seeder frame; planting depth controllers, such as linear actuators; measurement controllers, such as electric seed drive motors, hydraulic seed drive motors or row control clutches; hybrid selection controllers, such as seed metering drive motors, or other triggers programmed to selectively allow or prevent seed or an air-seed mixture from being delivered to or
Petition 870190123722, of 11/26/2019, p. 48/105
42/90 seed dispensers or central volume hoppers; measurement controllers, such as electric seed drive motors, or hydraulic seed drive motors; controllers of seed conveyor systems, such as controllers for a seed delivery conveyor motor; marker controllers, such as a controller for a pneumatic or hydraulic actuator; or rate controllers for pesticide applications, such as measurement trigger controllers, size or hole position controllers.
[073] In one embodiment, examples of sensors 112 that can be used with tillage equipment include position sensors for tools such as rods or discs; tool position sensors for such tools that are configured to detect depth, tool set angle or side spacing; downward force sensors; or drag force sensors. In one embodiment, examples of controllers 114 that can be used with tillage equipment include down force controllers or tool position controllers, such as
configured controllers to control depth in tool, angle of set in tools or spacing side. [074] In a modality , examples From 112 sensors what
can be used in relation to the appliance to apply fertilizer, insecticide, fungicide and more, such as seed fertilizer starter systems,
Petition 870190123722, of 11/26/2019, p. 49/105
43/90 subsoil fertilizer applicators or fertilizer sprayers, include: fluid system criteria sensors, such as flow sensors or pressure sensors; sensors indicating that spray head valves or fluid line valves are open; sensors associated with tanks, such as fill level sensors; sectional or total system supply line sensors, or row-specific supply line sensors; or kinematic sensors such as accelerometers arranged on spray booms. In one embodiment, examples of controllers 114 that can be used with such an apparatus include pump speed controllers; valve controllers that are programmed to control pressure, flow, direction, PWM and more; or position actuators, such as for boom height, subsoil plow depth or boom position.
[075] In one embodiment, examples of sensors 112 that can be used with harvesters include production monitors, such as strain plate strain gauges or position sensors, capacitive flow sensors, load sensors, weight sensors, or torque sensors associated with elevators or augers, or optical sensors or other electromagnetic grain height sensors; grain moisture sensors, such as capacitive sensors; grain loss sensors, including impact sensors, optical or capacitive; harvester head operating criteria sensors such as harvester head height, harvester head type, pallet plate clearance, feeder speed and
Petition 870190123722, of 11/26/2019, p. 50/105
44/90 reel speed; separator operating criteria sensors, such as concave opening, rotor speed, shoe clearance, or clearance sensors; auger sensors for position, operation or speed; or engine speed sensors. In one embodiment, examples of controllers 114 that can be used with harvesters include harvester head operating criteria controllers for elements such as harvester head height, harvester head type, pallet plate clearance, feeder speed or reel speed ; separator operating criteria controllers for features such as hollow opening, rotor speed, shoe clearance, or clearance; or controllers for auger position, operation or speed.
[07 6] In one embodiment, examples of sensors 112 that can be used with grain carts include weight sensors, or sensors for position, operation, or auger speed. In one embodiment, examples of controllers 114
that can be used with carts in grains include position controllers, operation or speed gimlet. [077] In one mode, examples From 112 sensors and
of controllers 114 can be installed in unmanned aerial vehicles (UAVs) or drones. Such sensors may include cameras with effective detectors for any range of the electromagnetic spectrum including visible, infrared, ultraviolet, near infrared (NIR) and more; accelerometers; altimeters; temperature sensors; humidity sensors; tube sensors
Petition 870190123722, of 11/26/2019, p. 51/105
45/90 pitot or other air speed sensors or wind speed sensors; battery life sensors; or radar emitters and apparatus for detecting reflected radar energy. Such controllers may include guidance or motor control devices, surface control controllers, camera controllers or controllers programmed to connect, operate, obtain data, manage and configure any of the sensors indicated above. Examples are disclosed in patent application US 14 / 831,165 and the present disclosure acknowledges that other patent disclosure.
[078] In one embodiment, sensors 112 and controllers 114 can be incorporated into a soil sampling and measurement device that is configured or programmed to sample soil and perform soil chemistry tests, soil moisture tests and other tests relative to the soil. For example, the apparatus disclosed in US patent 8,767,194 and US patent 8,712,148 can be used, and the present disclosure assumes knowledge of those patent disclosures.
[079] In one embodiment, sensors 112 and controllers 114 can comprise meteorological devices for monitoring field weather conditions. For example, the apparatus revealed in provisional application US 62 / 154,207, filed on April 29, 2015, in provisional application US 62 / 175,160, filed on June 12, 2015, in provisional application US 62 / 198,060, filed on 28 July 2015 and provisional order US 62 / 220,852, filed on September 18, 2015, can be used, and the present disclosure acknowledges these
Petition 870190123722, of 11/26/2019, p. 52/105
46/90 patent disclosures.
[080] 2.4 PROCESS OVERVIEW - AGRONOMIC MODEL TRAINING [081] In one embodiment, the agricultural intelligence computer system 130 is programmed or configured to create an agronomic model. In this context, an agronomic model is a data structure in the memory of the agricultural intelligence computer system 130 comprising field data 106, such as identification data and harvest data for one or more fields. The agronomic model can also comprise calculated agronomic properties that describe conditions that can affect the growth of one or more plantations in a field, or properties of one or more plantations, or both. In addition, an agronomic model can comprise recommendations based on agronomic factors such as planting recommendations, irrigation recommendations, planting recommendations and harvest recommendations. Agronomic factors can also be used to estimate one or more results related to plantation, such as agronomic yield. The agronomic yield of a plantation is an estimate of the amount of the plantation that is produced, or in some instances the revenue or profit obtained from the harvested crop.
[082] In one embodiment, the agricultural intelligence computer system 130 can use a pre-configured agronomic model to calculate agronomic properties related to location and plantation information currently received for one or more fields. The pre-configured agronomic model is based on field data
Petition 870190123722, of 11/26/2019, p. 53/105
47/90 previously processed, including, but not limited to, identification data, harvest data, fertilizer data and weather data. The pre-configured agronomic model may have been cross-validated to ensure model accuracy. Cross-validation may include a comparison with soil veracity that compares predicted results with actual results in a field, such as a comparison of estimated rainfall with a rain gauge or sensor providing meteorological data in the same or nearby location or an estimate of soil content. nitrogen with a soil sample measurement.
[083] Figure 3 illustrates a programmed process by which the agricultural intelligence computer system generates one or more preconfigured agronomic models using field data provided by one or more data sources. Figure 3 can serve as an algorithm or instructions for programming the functional elements of the agricultural intelligence computer system 130 to perform the operations that are now described.
[084] In block 305, the agricultural intelligence computer system 130 is configured or programmed to implement pre-processing of agronomic data from field data received from one or more data sources. Field data received from one or more data sources can be pre-processed for the purpose of removing noise and distortion effects within agronomic data including measured outliers that would predispose received field data values. Pre-processing modalities for agronomic data may include, but are not limited to,
Petition 870190123722, of 11/26/2019, p. 54/105
48/90 remove data values commonly associated with outlier data values, specific measured data points that are known to unnecessarily distort other data values, data smoothing techniques used to remove or reduce additive or multiplicative noise effects, and others filtering or data derivation techniques used to provide clear distinctions between positive and negative data inputs.
[085] In block 310, the agricultural intelligence computer system 130 is configured or programmed to perform subset selection of data using the pre-processed field data in order to identify data sets useful for generating the initial agronomic model. The agricultural intelligence computer system 130 can implement subset data selection techniques including, but not limited to, a genetic algorithm method, a model method for all subsets, a sequential search method, a regression method gradually, a particle multitude optimization method and an ant colony optimization method. For example, a genetic algorithm selection technique uses an adaptive heuristic research algorithm, based on evolutionary principles of natural selection and genetics, to determine and evaluate data sets within pre-processed agronomic data.
[086] In block 315, the agricultural intelligence computer system 130 is configured or programmed to implement field data set evaluation. In one embodiment, a specific field data set is
Petition 870190123722, of 11/26/2019, p. 55/105
49/90 assessed when creating an agronomic model and using specific quality thresholds for the agronomic model created. Agronomic models can be compared using cross-validation techniques including, but not limited to, root cross-validation of the mean square error of leaving one out (RMSECV), mean absolute error and mean percentage error. For example, RMSECV can cross-validate agronomic models by comparing predicted agronomic property values created by the agronomic model with historical agronomic property values collected and analyzed. In one embodiment, the agronomic data set assessment logic is used as a feedback loop where agronomic data sets that do not meet configured quality thresholds are used during future data subset selection steps (block 310).
[087] In block 320, the agricultural intelligence computer system 130 is configured or programmed to implement the creation of an agronomic model based on cross-validated agronomic data sets. In one embodiment, the creation of an agronomic model can implement multiple variable regression techniques to create pre-configured agronomic data models.
[088] In block 325, the agricultural intelligence computer system 130 is configured or programmed to store pre-configured agronomic data models for future evaluation of field data.
[089] 2.5 IMPLEMENTATION EXAMPLE - HARDWARE OVERVIEW [090] According to one modality, the techniques
Petition 870190123722, of 11/26/2019, p. 56/105
50/90 described in this document are implemented using one or more special-purpose computing devices. Special-purpose computing devices can be physically connected to perform the techniques, or they can include digital electronic devices such as one or more application-specific integrated circuits (ASICs) or field programmable port arrays (FPGAs) that are persistently programmed to perform the techniques, or may include one or more general purpose hardware processors programmed to perform the techniques according to program instructions in firmware, memory, other storage, or in a combination. Such special-purpose computing devices can also combine physically connected connected logic, ASICs, or FPGAs with customized programming to execute the techniques. Special-purpose computing devices can be desktop computer systems, portable computer systems, portable devices, network devices, or any other device that incorporates physically connected and / or program logic to implement the techniques.
[091] For example, figure 4 is a block diagram illustrating a computer system 400 in which a modality of the invention can be implemented. Computer system 400 includes a bus 402 or other communication mechanism for transmitting information, and a hardware processor 404 coupled to bus 402 for processing information. The hardware processor 404 can be, for example, a general purpose microprocessor.
[092] Computer system 400 also includes main memory 406, such as access memory
Petition 870190123722, of 11/26/2019, p. 57/105
51/90 random (RAM) or other dynamic storage device, coupled to bus 402 to store information and instructions to be executed by processor 404. Main memory 406 can also be used to store temporary variables or other intermediate information during instructions execution to be executed by the 404 processor. Such instructions, when stored on non-transitory storage media accessible to the 404 processor, render the computer system 400 on a special-purpose machine that is customized to perform the operations specified in the instructions.
[093] Computer system 400 additionally includes a read-only memory (ROM) 408 or other static storage device coupled to bus 402 to store static information and instructions for processor 404. A storage device 410, such as a disk magnetic, optical disk or solid state drive, is provided and coupled to the 402 bus to store information and instructions.
[094] 0 system in computer 400 can be coupled via bus 402 a one dial 412, such as a pipe in cathode rays ( CRT ), to display information for one
computer user. An input device 414, including an alphanumeric keyboard and other keys, is coupled to bus 402 to input information and command selections for the 404 processor. Another type of user input device is the cursor control 416, such as a mouse, a stationary mouse, or cursor direction keys to enter direction information and
Petition 870190123722, of 11/26/2019, p. 58/105
52/90 command selections for processor 404 and to control cursor movement on display 412. This input device typically has two degrees of freedom on two axes, a first axis (for example, x) and a second axis (for example, y), which allow the device to specify positions on a plane.
[095] Computer system 400 can implement the techniques described in this document using custom built-in logic, one or more ASICs or FPGAs, firmware and / or program logic that in combination with the computer system induce or program the computer system 400 to be a special use machine. According to one embodiment, the techniques in this document are performed by computer system 400 in response to processor 404 executing one or more sequences of one or more instructions contained in main memory 406. Such instructions can be read in main memory 406 of another storage media, such as storage device 410. Execution of instruction sequences contained in main memory 406 induces processor 404 to perform the process steps described in this
document. In modalities alternatives, set in circuits embedded can be used in place or in combination with instructions for software. [096] 0 term media storage as used
this document refers to any non-transitory media that store data and / or instructions that induce a machine to operate in a specific mode. Such storage media may comprise non-volatile media and / or volatile media. Non-volatile media include,
Petition 870190123722, of 11/26/2019, p. 59/105
53/90 eg optical discs, magnetic discs or solid state drives such as the storage device 410. Volatile media include dynamic memory, such as main memory 406. Common forms of storage media include, for example, a floppy disk , a floppy disk, hard drive, solid state drive, magnetic tape, or any other magnetic data storage media, a CD-ROM, any other optical data storage media, any physical media with hole patterns, a RAM , a PROM and EPROM, a FLASHEPROM, NVRAM, any other chip or memory cartridge.
[097] Storage media are distinct from transmission media, but can be used in combination with them. Transmission media participate in the transfer of information between storage media. For example, transmission media include coaxial cables, copper wire and optical fibers, including the wires that comprise the 402 bus. Transmission media can also take the form of acoustic or light waves, such as those generated during data communications. radio wave and infrared.
[098] Various forms of media can be involved by loading one or more strings of one or more instructions into the 404 processor for execution. For example, instructions can be loaded initially onto a magnetic disk or solid state drive on a remote computer. The remote computer can load the instructions into its dynamic memory and send the instructions over a telephone line using a modem. A local modem for computer system 400 can receive data over the line
Petition 870190123722, of 11/26/2019, p. 60/105
54/90 phone and use an infrared transmitter to convert the data into an infrared signal. An infrared detector can receive the data loaded in the infrared signal and an appropriate circuitry can place the data on bus 402. Bus 402 loads the data into main memory 406, from which processor 404 retrieves and executes instructions. Instructions received by main memory 406 can optionally be stored in storage device 410 before or after execution by processor 404.
[099] Computer system 400 also includes a communication interface 418 coupled to bus 402. Communication interface 418 provides a bidirectional data communication coupling for a network link 420 that is connected to a local network 422. For example , the communication interface 418 can be an integrated services digital network card (ISDN), cable modem, satellite modem or a modem to provide a data communication connection for a corresponding type of telephone line. As another example, communication interface 418 can be a local area network (LAN) card to provide a data communication connection to a compatible LAN. Wireless links can also be
implemented. In any such implementation, the interface of communication 418 send and receives signals electrical, electromagnetic or optical what carry streams of data
representing various types of information.
[0100] The network link 420 typically enables data communication through one or more networks to
Petition 870190123722, of 11/26/2019, p. 61/105
55/90 other data devices. For example, the network link 420 may provide a connection via the local network 422 to a host computer 424 or to data equipment operated by an Internet Service Provider (ISP) 426. ISP 426 in turn provides network services. data communication through the worldwide packet data communication network now commonly referred to as the Internet 428. The local network 422 and the Internet 428 use electrical, electromagnetic or optical signals that carry digital data streams. Signals through the various networks and signals on network link 420 and through communication interface 418, which carry digital data to and from computer system 400, are exemplary forms of transmission media.
[0101] Computer system 400 can send messages and receive data, including program code, via the network (s), the network link 420 and the communication interface 418. In the example of the Internet, a server 430 can transmit a requested code to an application program via the Internet 428, ISP 426, local network 422 and the communication interface 418.
[0102] The received code can be executed by processor 404 as it is received and / or stored in storage device 410, or other non-volatile storage, for later execution.
[0103] 3. AGRONOMIC NEURAL NETWORK [0104] Figure 7 represents a neural network architecture for computing one or more predicted yield values from one or more plantation-related entries. Neural networks, as described in this
Petition 870190123722, of 11/26/2019, p. 62/105
56/90 document, refer to a plurality of equations, each of which is configured to increase a plurality of inputs to produce an output. Individual weights are applied to each equation and increased as the neural network is trained. Thus, the greater the amount of training information, the stronger the neural network becomes to produce accurate outputs.
[0105] Individual neural networks are described in this document at high levels of generality based on inputs, outputs and type of neural network. A person of common knowledge in the art having data on inputs, outputs and type of neural network would be able to build a work modality using open source code. For example, open source software can be used to implement a particular architecture of a neural network such as the Visual Geometry Group (VGG) neural network created by the Department of Engineering Science at the University of Oxford. Open source software to implement the VGG neural network is available from TensorFlow, an open source software library developed by the Google Brain Team.
[0106] In figure 7, the deep neural network 700 includes a plurality of inputs that can be used to train an agronomic neural network, including plantation identification data 704, environmental data 708, management practice data 716 and additional data 720. Although figure 7 represents a plurality of neural networks trained in isolation, in one mode each individual neural network is trained as part of a cohesive system. Thus, the input data
Petition 870190123722, of 11/26/2019, p. 63/105
57/90
704, 708, 712, 716 and 720 can be used to train an individual neural network or to train a series of neural networks, as shown in Figure 7, which are then used to train a master neural network 740.
[0107] 3.1. TRAINING DATA [0108] Training data in general refers to the data sets that are used to train individual neural networks, the deep neural network 700, or a combination of the two. Each data set corresponds to the growth cycle of one or more plantations in one or more fields. Each data set used in training the deep neural network 700 includes at least one yield value for one or more plantations. Yield values are further described in this document, but they generally correspond to a value representing the completion of a growth cycle. Thus, an appropriate training data set corresponds to a plantation that has been planted, matured and harvested.
[0109] Training data also includes one or more of the plantation identification data 704, environmental data 708, management practice data 716 and additional data 720. Although the best training data sets include each type of data, individual neural networks can still be trained using some data sets that do not understand each type of data. For example, data sets comprising plantation identification data 704 and environmental data 708 would be useful in capturing the cross-section of effects on total yield of the type of plantation and the
Petition 870190123722, of 11/26/2019, p. 64/105
58/90 environment, while data comprising effects on the total performance of management practices would be lost. Thus, as with any neural network, the more complete the training data, the better the performance of the neural network.
[0110] In one mode, individual neural networks are run on incomplete data sets, that is, data sets that lose one or more of the types of data used to train the deep neural network 700. In a mode, when a set of Incomplete data is used, missing data types are assumed to be kept constant. Thus, in the example indicated above, average values can be used for management practice data 716 when training the deep neural network using a data set comprising plantation identification data 704 and environmental data 708.
[0111] Datasets can be provided using external data 110 and / or field data 106. For example, several plantation studies may contain data regarding types of plantations, weather events, management practices and additional information . Data from plantation studies can be combined into a plurality of data sets to train the deep neural network 700. In addition, data sets can be provided by field manager computing devices 104. Individual farmers can send data regarding their fields for the agricultural intelligence computer system 130. The agricultural intelligence computer system 130 can track additional field data for individual farmers. Per
Petition 870190123722, of 11/26/2019, p. 65/105
59/90 example, the agricultural intelligence computer system 130 can receive temperature and precipitation data from one or more sensors. Once a field has been harvested, the field manager computing device 104 can send production data to the agricultural intelligence computer system 130. Using the tracked and received data, the agricultural intelligence computer system 130 can generate a new data set to train the 700 deep neural network.
[0112] 3.2. NEURAL PLANTATION IDENTIFICATION NETWORK [0113] In one embodiment, a first neural network is trained using plantation identification data 704. Plantation identification data 704 can include input used to distinguish a type of crop in a different type of crop. plantation. For example, plantation identification data may include the relative maturity of a seed that is planted in a field. Plantation identification data may also include indicators for a type of crop. For example, a first value in a set of values can refer to a type of crop, such as corn, cotton or wheat, while a next series of values identifies the individual species of the plantation, such as the relative maturity of the seed .
[0114] In one embodiment, plantation identification data 704 includes the crop genotype 706. The crop genotype 706 can comprise a plurality of values that identify the genotype of the particular crop that is planted in the field or the genome of the crop that is planted in the field. The culture genotype can be
Petition 870190123722, of 11/26/2019, p. 66/105
60/90 compressed into a vector of values representing a DNA sequence from the plantation. Additionally or alternatively, the culture genotype 706 may comprise one or more single nucleotide polymorphisms (SNPs). Although the genome for a particular culture may be in the range of 50,000 values, using one or more SNPs allows a similar amount of information to be compressed into a smaller number of values.
[0115] In one embodiment, the culture genotype 706 may include one or more parts of a culture genome. For example, scientific research on genomes for maize can indicate which parts of the genome are most directly correlated with differences in plantation yield. Instead of using the entire maize genome to train and use the plantation identification neural network, the plantation identification neural network can be constructed using only those parts of the genome sequence that have been determined to have a correlation with plantation yield. Thus, when planting information is received for a particular field, the agricultural intelligence computer system 130 can be programmed or configured to identify particular parts of the crop genome based on parts of the crop genome that were used to train the neural network plantation identification.
[0116] In one embodiment, the deep neural network 700 includes a recurrent neural network (RNN) to compute effects on planting yield from plantation identification data 704. For example, an RNN can be trained using crop genomes as input and
Petition 870190123722, of 11/26/2019, p. 67/105
61/90 values of income as output. As RNN is trained, RNN can learn about which elements of the culture genome have an effect on yield and which elements of the culture genome do not have an effect on yield. Example RNN modalities for computing yield effects based on plantation identification data include neural networks of long and short term memories (LSTMs) and neural networks of recurrent door units (GRUs). The use of LSTMs or GRUs allows the neural network to retain information regarding effects on plantation yield regardless of when the information was introduced into the neural network.
[0117] In one embodiment, the RNN is used to compute an embedded genotype 730 incorporation. The incorporation layers, as described in this document, refer to the intermediate layers that contain information relevant to the effects of data entry on the yield of plantation. These embedding layers allow the master neural network to run on a smaller amount of data, thereby improving the performance of the master neural network by reducing the processing capacity to train and run the master neural network and reducing the time spent training and run the master neural network. In the case of plantation identification data 704, the RNN encodes effects on planting yield of the crop genotype into a smaller number of values that can later be used in a master neural network to compute one or more yield values.
[0118] Although each incorporation is described as an encoding of an effect on total income based on
Petition 870190123722, of 11/26/2019, p. 68/105
62/90 in particular data, each individual incorporation is not coded to be a device by itself. Each incorporation is configured during the training stage to include useful information regarding input data that can have an effect on total performance. The information included in a first incorporation is designed to interact with information included in a second incorporation. For example, certain crop genotypes may be more resistant to high temperatures than other crop genotypes. Thus, the incorporation of the discovered genotype 730 may contain information that identifies how strongly the effects of high temperatures on the discovered environmental incorporation will influence the total plantation yield.
[0119] 3.3. ENVIRONMENTAL NEURAL NETWORK [0120] In one modality, a second neural network is trained using environmental data 708. Environmental data 708 can include any information about the environment during a plantation growing season. For example, environmental data 708 may include information regarding weather conditions, such as temperature, precipitation, amount of snow, wind speed and occurrence of abnormal events such as hurricanes, floods and earthquakes. In addition or alternatively, environmental data 708 may include properties for each field, such as topography and soil properties. In one embodiment, environmental data is divided into one or more of temporal environmental data, spatial environmental data or spatiotemporal environmental data.
Petition 870190123722, of 11/26/2019, p. 69/105
63/90 [0121] 3.3.1. TEMPORARY ENVIRONMENTAL DATA [0122] Temporal environmental data 710 includes information about the environment that changes over time, but is treated as being consistent over an entire field. The temporal environmental data 710 can be stored as one or more time series, that is, a vector of values in which each represents a measurement or estimate at a particular time in a growing season. Time series can include values for each minute, hour, day and / or week depending on the time series. For example, in the case of precipitation, hourly precipitation levels may be more useful for covering small convective storms while daily temperature levels can be used to minimize the amount of data stored and used in training and processing neural networks. Temporal environmental data 710 may include, but is not limited to, either daily temperature, daily precipitation level, daily snow quantity level, or abnormal occurrence or non-daily occurrence.
[0123] In one embodiment, temporal environmental data 710 may comprise a plurality of time series for a particular type of data. For example, three (3) time series can be used for temperature for each field: a higher temperature time series, a lower temperature time series and an average temperature time series. The highest temperature time series, for example, can include a series of data values representing the highest temperature as measured by a thermometer in a field
Petition 870190123722, of 11/26/2019, p. 70/105
64/90 for each day of a growing season starting on the planting date.
[0124] In one embodiment, the deep neural network 700 includes a recurrent neural network (RNN) to compute effects on planting yield from temporal environmental data 708. For example, an RNN can be trained using time series as input and yield values as an exit. Each time series can be identified as a different input for a single convolution neural network. Alternatively, each time series can be used as input to its own RNN. The resulting yield effects of each of the time series can then be used as inputs to an environmental temporal neural network in order to compute a total yield effect of a plurality of different time series. Example RNN modalities for computing yield effects based on temporal environmental data include neural networks of long and short term memories (LSTMs) and neural networks of recurrent port units (GRUs).
[0125] 3.3.2. SPATIAL ENVIRONMENTAL DATA [0126] Spatial environmental data 712 includes information about the environment that changes spatially, but which is treated as being consistent over time. Spatial environmental data 712 can be stored as one or more stacked spatial layers, that is, arrays of values each representing a measurement or estimate at a particular location in the field. For example, each element of a first matrix can represent a single pixel five meters by five meters
Petition 870190123722, of 11/26/2019, p. 71/105
65/90 which corresponds to a location in the field. Thus, the first element of the matrix can comprise a measurement or estimate in the upper left of five meters by five meters from the field. The pixel size for different types of spatial data may differ, however the pixel size for a particular type of data may remain constant for each entry of the same data type. For example, topography maps of the field may be available at higher resolutions than maps of soil properties of the field. Thus, the spatial resolution between topography maps and maps of soil properties may differ. Also, each topography map for each field of training data can use the same pixel size.
[0127] In one embodiment, spatial environmental data 712 comprises a plurality of spatial maps for a particular type of data. For example, soil property maps can include a first map for soil sand content, a second map for soil silt content, and a third map for soil clay content. Additional spatial maps may include, but are not limited to, the soil's nitrogen content, soil type, and organic carbon content in the soil. Spatial environmental data 712 may additionally include different layers of soil properties. For example, organic carbon content in the soil can be measured in an upper layer and in one or more layers of subsurface soil.
[0128] Where spatial elements are assumed to change over time, a plurality of spatial environmental maps can be used as space data
Petition 870190123722, of 11/26/2019, p. 72/105
66/90 storms. For example, soil moisture content can change daily based on water application, precipitation, temperature and water absorption by the plantation. A plurality of moisture content maps can be used as input data for a neural network, where each moisture content map identifies moisture content in a plurality of locations at a different point in the growing season.
[0129] In one embodiment, the deep neural network 700 contains a convolutional neural network (CNN) to compute effects on the yield of spatial environmental data. A convolutional neural network is capable of learning from image data, such as spatial environmental data 712. As each data point represents a pixel, the convolutional neural network can track effects of pixels close together and map an effect on the yield from various spatial maps. As with recurrent neural networks, each spatial map can correspond to its own convolutional neural network, each spatial map can be used as a different input for a comprehensive neural network, or any combination of spatial maps can be used as different inputs for a convolutional neural network. For example, in the case of spatio-temporal data, each map of the plurality of maps can be used as different inputs for the convolutional neural network.
[0130] 3.3.3. DISCOVERED ENVIRONMENTAL INCORPORATION [0131] In one embodiment, the deep neural network uses one or more RNNs for temporal environmental data 710 and one or more CNNs for spatial environmental data 712
Petition 870190123722, of 11/26/2019, p. 73/105
67/90 to generate the discovered environmental incorporation 732. The discovered environmental incorporation identifies a cross-section between temporal environmental factors and spatial environmental factors. As with the incorporation of discovered genotype 730, the environmental incorporation discovered 732 encodes relevant information for a total effect on plantation yield based on all environmental data.
[0132] Although figure 7 represents a single stage of incorporation for environmental data, in one modality each CNN and RNN produces an incorporation for a particular type of data. For example, a first incorporation based on daily precipitation vectors computed through an RNN can encode effects of daily precipitation in total yield while a second incorporation based on topography maps computed through a CNN can encode effects of topography in yield total.
[0133] Each of the individual incorporations can retain its individual properties. For example, time series incorporations can retain the properties of linear vectors while spatial map incorporations can retain the properties of matrices. Each incorporation can be used as an input into a neural network of environmental data that produces a combined incorporation of environmental effects into total plantation yield. The environmental incorporation discovered 732, as a combination of time series data and spatial data, can be three-dimensional. For example, the discovered 732 environmental incorporation can be stored
Petition 870190123722, of 11/26/2019, p. 74/105
68/90 as a three-dimensional matrix in which two dimensions represent space and the third dimension represents time.
[0134] 3.4. NEURAL MANAGEMENT PRACTICE NETWORK [0135] In one embodiment, a third neural network is trained using the 716 management practice data. The 716 management practice data can include information regarding the management practices of one or more fields. For example, data on management practices may include crop information, including crop type, seed depth and planting population, nutrient application, nutrient inhibitor application, water application and any other management practices relating to one or more more field.
[0136] Management practice data 716 can be stored as maps as applied 718. Maps as applied 718 can be stored as one or more stacked spatial layers, that is, matrices of values in which each represents a measurement or estimate at a particular location in the field. Additionally, one or more added values can indicate application times. For example, a fertilizer map as applied may contain a plurality of pixels, each of which identifies, for a particular area of the field, an amount of fertilizer that has been applied. The fertilizer map as applied may additionally contain one or more pixels that identify the moment of application. For example, a first matrix value representing the map as applied may indicate a number of days in the growing season when the fertilizer was added. The rest of the matrix can start
Petition 870190123722, of 11/26/2019, p. 75/105
69/90 in the next row and column to represent the spatial application of the fertilizer.
[0137] Additionally or alternatively, some moment information can be stored as a separate time series. For example, a first time series can indicate which days included a field water application. A corresponding matrix can indicate quantities of applications across the field. Separating the moment information from the application quantity information allows the deep neural network 700 to identify effects of applying water at different times as well as effects of applying water at different rates.
[0138] In one embodiment, the deep neural network 700 includes a CNN to compute effects on planting yield from management practice data 716. For example, a CNN can be trained using data from management practices such as inputs and yields from plantations as exits. As with recurrent neural networks, each spatial map can correspond to its own convolutional neural network, each spatial map can be used as a different input for a comprehensive neural network, or any combination of spatial maps can be used as different inputs for a convolutional neural network.
[0139] In one embodiment, CNN is used to compute a 736 discovery management practice incorporation. The 736 discovery management practice incorporation encodes relevant information for a total effect on plantation yield based on all individual management practices. . Thus, the incorporation of practice
Petition 870190123722, of 11/26/2019, p. 76/105
70/90 of discovery management 736 includes combinations of the effects of individual management practices, such as population maps as applied and seed depth maps as applied.
[0140] Although figure 7 represents a single incorporation stage for management practice data, in one modality each CNN produces an incorporation for a particular type of data. For example, a first incorporation based on fertilizer application computed through a CNN can encode effects of fertilizer application in full yield while a second incorporation based on planting population computed through a CNN can encode effects of the planting population in total yield. Each incorporation can be used as an input into a neural network of management practice data that produces a combined incorporation of management practice effects into total plantation yield.
[0141] 3.5. ADDITIONAL NEURAL NETWORKS [0142] In one embodiment, a fourth neural network is trained using the additional data 720. The additional data 720 can include any additional information that may have a connection to the total plantation yield by itself or in combination with any one of the previous data sources. Two examples of additional data 720 include satellite images 722 and past performance maps 724.
[0143] Satellite images 722 comprise one or more images of a particular field from one or more satellites. Satellite images 722 may contain several
Petition 870190123722, of 11/26/2019, p. 77/105
71/90 stacked layers of satellite images taken at the same time, but at different frequencies. For example, a first layer can correspond to a blue frequency, a second layer can correspond to a red frequency and a third layer can correspond to an infrared frequency. Thus, for a single instance of satellite images 722, three different arrays can be used. Additionally or alternatively, a three-dimensional matrix can represent the stacked layers of satellite images.
[0144] Satellite images 722 may correspond to particular times or periods of the growing season. For example, satellite images 722 may include images taken from the field at planting time and on a particular number of days in the growing season, such as a hundred days in the growing season. As another example, satellite images can be obtained from fields periodically, such as every week. By keeping satellite images consistent across different fields, the deep neural network 700 is able to properly weight the effects of variances on satellite images. For example, a satellite image taken immediately after planting is likely to look very different from a satellite image taken immediately before the last application of nitrogen by means of side preparation.
[0145] In one embodiment, data on each of the 722 satellite images can identify when the satellite image was produced within the growing season. For example, a first value in an array can identify
Petition 870190123722, of 11/26/2019, p. 78/105
72/90 a number of days after planting when the satellite image was generated. The values representing the satellite image can start in a new row and column of the matrix. Alternatively, data identifying when satellite images were obtained can be stored in a separate time series.
[0146] In one embodiment, satellite images 722 are produced on a similar scale for each field. If satellite images are not scaled properly in relation to each other, certain satellite images may appear to have a different effect on yield than equivalent satellite images at a different resolution. Thus, a standardized resolution can be used for each satellite image. Image interpellation techniques can be used on images of varying resolutions to standardize image resolution.
[0147] Past performance maps 724 comprise past performance maps for each set of fields. Thus, for a particular data set, a past yield map can identify yields from field plantations for previous years. Past yield maps 724 can identify plantation yield in particular resolutions. For example, each pixel on a past yield map can identify a planting yield for a space of 5 meters by 5 meters in the field. Thus, past performance maps 724 can be represented by a matrix of values. As with satellite images 722, past performance maps can be stored in a resolution consistent with each other.
Petition 870190123722, of 11/26/2019, p. 79/105
73/90 [0148] In one embodiment, past performance maps 724 identify, for each location, one or more different performance values. For example, a first past yield map can identify total crop yield for each location. A second map of past earnings can identify a total profit for each location. A third map of past yields can identify one or more crop values, such as protein quality for each location.
[0149] In one embodiment, the deep neural network uses one or more CNNs for satellite images 722 and one or more CNNs for past performance maps 724 to generate the incorporation of additional data 738. The incorporation of additional data identifies a cross-section between any additional data values, such as satellite images and past performance maps. As with the other incorporations discussed in this document, the incorporation of additional data 738 encodes information relevant for total effects on plantation yield based on all additional data.
[0150] Although figure 7 represents a single incorporation stage for additional data, in one embodiment each CNN produces an incorporation for a particular type of data. For example, a first incorporation based on satellite images obtained at a particular stage of plantation development computed using a CNN can encode correlations between satellite images and total yield while a second incorporation based on past yield maps computed by
Petition 870190123722, of 11/26/2019, p. 80/105
74/90 means of a CNN can encode effects of past income values into current total income values. In addition, individual types of data can be processed through one or more CNNs. For example, satellite images can be processed using a single CNN or through a CNN for each frequency image.
[0151] 3.6. INCORPORATIONS OF INTERMEDIATE TRANSVERSAL CUTS [0152] In one embodiment, the deep neural network 700 contains one or more incorporations of intermediate cross sections. Incorporations of intermediate cross sections encode relationships between two or more different types of data. Incorporations of intermediate cross sections can be useful outside the deep neural network 700 when comparing plantations in different fields. For example, two plantations in different parts of the world with the same cross-sectional plantation and environmental incorporations will respond similarly to the same management practices despite the fact that the two plantations may be different and may have been exposed to weather conditions. many different.
[0153] An example of an intermediary cross-sectional incorporation is the 734 plantation and environmental cross-sectional incorporation. The 734 plantation and environmental cross-sectional incorporation encodes information relevant to the effects of a combination of plantation type and environmental conditions in the plantation. The cross-sectional incorporation of plantation and environmental 734 can be trained using the incorporation of genotype
Petition 870190123722, of 11/26/2019, p. 81/105
75/90 discovered 730 and the environmental incorporation discovered 732 as inputs.
[0154] Although figure 7 represents a limited number of cross-section incorporations, modalities can comprise any number of cross-section incorporations. For example, a modality may not include cross-sectional incorporation. In a mode like this, each type of data is used as a different input to a master neural network. In another embodiment, additional incorporations are included in figure 7. For example, the 734 plantation and environmental cross-sectional incorporation can be combined with the discovery management practice incorporation 736 to create a cross-section between the two incorporations. These two incorporations can be combined with the addition of additional data 738 in order to create a single incorporation used for introduction into the master neural network 740.
[0155] 3.7. MASTER NEURAL NETWORK [0156] In one embodiment, a master neural network 740 transforms a plurality of different inputs into one or more yield values 750. The master neural network can comprise a CNN with one or more layers of neural network 742. In one In this mode, the master neural network is configured to accept as input at least plantation identification data 704, environmental data 708 and data on management practices 716.
[0157] Alternatively, the master neural network 740 can be configured to accept as input one or more incorporations of information relevant to plantation yield. The use of encoded information in incorporations
Petition 870190123722, of 11/26/2019, p. 82/105
76/90 significantly reduces the amount of data used to train or execute the master neural network 740 without significantly reducing the quality of the master neural network 740. In one embodiment, a single layer of incorporation is used as input to the master neural network 740. The single layer of incorporation can encode all information relevant to planting yield for each type of data. Additionally or alternatively, the master neural network 740 can be configured to use a plurality of intermediate layers as an input.
[0158] In one embodiment, each of the layers of neural network 742 comprises a plurality of nodes configured to transform the inputs into an output value. In embodiments where the layers of neural network 742 comprise a plurality of layers, outputs from each node can be used as inputs for each node in subsequent layers. Neural network layers 742 may also comprise a plurality of weights, each of which is associated with an individual node. The agricultural intelligence computer system 130 can use standard machine learning techniques to update weights over time. For example, the agricultural intelligence computer system 130 can update weights based on training data by minimizing the difference between training data outputs and actual yield values in training data.
[0159] 3.8. INCOME VALUES [0160] In one modality, the master neural network 740 is configured to produce one or more yield values 750. The yield values 750 correspond to
Petition 870190123722, of 11/26/2019, p. 83/105
77/90 results of harvesting a plantation. Yield values 750 may include one or more of total yield 752, risk adjusted yield 754, plantation quality values 756 and total profits 758. One or more of yield values 750 can be trained as outputs for the master neural network 740. For example, the master neural network 740 can be trained using the total throughput 752 as a single output. Additionally or alternatively, the master neural network 740 can be trained to produce an output comprising a plurality of values. For example, a vector output may include a first value for total yield 752, a second value for one or more
plantation 756 and a third value To the profits totals 758.[0161] 0 total yield 752 refers to to a amount in total production of a plantation to an area
particular. For example, total yield 752 can be measured in bushels of the plantation per acre of land. The total yield 752 can be converted into an absolute yield by computing a product of the total yield 752 over the area of a relevant field. Thus, when the neural network is run on current data relating to a farmer's field, the neural network can be used to compute the total amount of a plantation that the farmer's field will produce.
[0162] The risk-adjusted yield 754 refers to a variability in the expected result. Risk adjusted yield 754 can be a single value, such as an expectation value, or a range of values identifying
Petition 870190123722, of 11/26/2019, p. 84/105
78/90 a probabilistic distribution of plantation yield. The risk-adjusted yield 754 can take into account uncertainty in the neural network and / or in the input data. For example, training data for weather conditions can be relatively accurate as precipitation and temperature levels are measured directly. On the other hand, meteorological data may be less accurate when the data is run on a particular data set. For example, if the neural network is used to determine which crops and / or management practices should be used before planting, then the weather conditions during the growing season may not be known. Instead, the neural network can be run based on predictions of weather conditions, such as temperature and precipitation forecasts. As predictions are merely predictions, each prediction can be associated with an uncertainty, thereby creating a range of possible values of weather conditions to be introduced into the neural network. The master neural network 740 can compute a risk-adjusted yield 754 by taking into account uncertainty in weather forecasts, such as when running the neural network with a range of inputs or several times with sampled inputs from a distribution of probable condition scenarios weather.
[0163] Plantation quality values 756 refer to one or more values describing the quality of a plantation produced. Examples of 756 planting quality values include nutrient values, such as protein content, physical values such as weight, strength, hardness, vitreous quality and color, and
Petition 870190123722, of 11/26/2019, p. 85/105
79/90 widespread qualities such as grain quality, wheat quality, rice quality and cotton quality. Generalized quality values can be defined commercially. For example, cotton quality can depend on a variety of factors including strength, fiber length and color.
[0164] Different planting quality values 756 can be used in training data as outputs of the deep neural network 700. For example, output training data for the deep neural network 700 may include a vector comprising values for protein content wheat, hardness, color, moisture content and weight. By using the plantation quality values 756 as an output, the deep neural network 700 is able to provide more information to a farmer than exactly the amount of yield that a field can produce. Plantation quality values 756 allow a farmer to determine whether a particular field is ideal for planting a particular crop. For example, a field may be able to produce more corn than wheat, but of much lower quality. In a case like this, a farmer may wish to plant higher quality wheat than lower quality corn.
[0165] Total profits 758 refer to profits from the sale of a crop produced in a field. As with total income 752, total profits can be based on a particular area, such as dollars per acre, or added to an absolute profit figure. In one embodiment, the deep neural network 700 is trained using actual total profits from past crop sales. In other modalities, the network
Petition 870190123722, of 11/26/2019, p. 86/105
80/90 deep neural 700 is trained using one or more other yield values, such as total yield 752, risk adjusted yield 754 and / or planting quality values 756. When deep neural network 700 is run on In a new data set, the agricultural intelligence computer system 130 can be programmed or configured to compute expected values for total profits 758 based, at least in part, on one or more other income values.
[0166] Expected values for total profits 758 can be computed based on the expected value of the plantation, the total yield of the plantation and the cost of producing the plantation. Thus, an example equation can comprise P = Y * E - C where P is the yield of the plantation, E is the expected value of the plantation and C is a cost of production.
[0167] Expected plantation value can be computed based on past crop sales to the plantation. For example, the expected value of the plantation may be the average selling price of the crops over the past five years. As another example, the expected value of the plantation can be computed based on trends in crop sales. The cost of production may include seed costs, fertilization costs, estimated labor costs, irrigation costs and any additional costs associated with planting, managing and harvesting a plantation.
[0168] Expected values for total profits 758 can additionally factor into one or more values of 756 plantation quality. For example, expected values of plantation can be based on plantation qualities
Petition 870190123722, of 11/26/2019, p. 87/105
81/90 different. Thus, a plurality of expected values can be identified for a particular type of plantation based on different plantation qualities. The expected value can be based on crop sales over a previous period of time for different planting quality values. For example, sales of past crops can be correlated with protein content for a particular type of crop. Thus, the expected values for total profits 758 of the particular plantation type can be computed based on a computed protein content for a particular plantation type plantation.
[0169] 4. APPLICATIONS [0170] 4.1. IMPLEMENTATION OF THE NEURAL NETWORK [0171] Figure 8 represents an example method for executing an agronomic neural network. A server computer system can execute instructions to execute the neural network in response to receiving a request for estimated yield values based on one or more entries.
[0172] In step 802, a particular data set relating to one or more agricultural fields is received in a server computing system, in which the particular data set comprises identification data for particular plantations, particular environmental data and data of particular management practices. The particular data set can be received from a single source or from a plurality of sources. For example, the server computing system can receive first data from the field manager computing device if
Petition 870190123722, of 11/26/2019, p. 88/105
82/90 relating to management practices for a field and second data from an external server computer relating to the effects of weather conditions on the field.
[0173] Plantation identification data can be
received directly of a device computing manager in fields For example, one user can identify , per middle of an application running on a device in manager computing in fields, one
culture that the user wants to plant. In one mode, the user additionally selects a particular type of seed for planting. The server computer system can store data identifying genome sequences, SNPs and / or parts of genome sequences for each type of seed. Additionally or alternatively, the server computer system can request sequences of genomes, SNPs and / or parts of genome sequences for seed types identified by a field manager computing device.
[0174] Environmental data for the particular data set can be received from any one of a plurality of sources. Soil data for a field can be received directly from a field manager computing device or from one or more soil tests performed in the field. Additionally or alternatively, soil data can be received from one or more external databases such as the Geographic Database of Soil Surveys (SSURGO). Weather data can be received from one or more external servers configured to produce weather forecasts for
Petition 870190123722, of 11/26/2019, p. 89/105
83/90 the future. Additionally or alternatively, meteorological data can be computed based on meteorological data passed to a particular field. For example, a server computer system can compute average temperatures and rainfall for a field over the past five years using previously received temperature and precipitation measurements.
[0175] The field manager computing device can also send management practice data to the server computer system. For example, a user can identify, by means of an application running on a field manager computing device, planned crop information, including crop type, seed depth and planting population, planned nutrient application, inhibitor application planned nutrient, planned water application and / or any other planned management practices.
[0176] In one embodiment, the server computer system stores data relating to one or more agricultural fields. For example, the server computer system can receive data from each year identifying an income for a particular field. The server computer system can store the data as past performance maps. Thus, a first part of the private data set can be stored on the server computer system, a second part can be received from an external server computer and a third part can be received from a field manager computing device.
Petition 870190123722, of 11/26/2019, p. 90/105
84/90 [0177] In step 804, a crop identification effect in yield for the one or more agricultural fields is computed from the identification data of particular plantations using a first neural network configured using plantation identification data as input and plantation yield data as an output. For example, the deep neural network 700 can generate a genotype incorporation discovered for the one or more agricultural fields through a recurrent neural network that accepts culture genotype as input.
[0178] In step 806, an environmental effect on yield for the one or more agricultural fields is computed from the particular environmental data using a second neural network configured using environmental data as input and plantation yield data as an output. For example, the deep neural network 700 can generate an environmental incorporation discovered for the one or more agricultural fields through a recurrent neural network that accepts temporal environmental data as input and a convolution neural network that accepts spatial environmental data as input.
[0179] In step 808, an effect of yield management practice for the one or more agricultural fields is computed from the data of particular management practices using a third neural network configured using management practice data such as input and data from income from plantations as a way out. For example, the deep neural network 7 00 can generate an incorporation of discovered management practice for the one or more agricultural fields through a neural network of
Petition 870190123722, of 11/26/2019, p. 91/105
85/90 convolution that accepts data from management practices as input.
[0180] In modalities, the deep neural network 700 is executed with additional data, such as maps of yields passed to the particular field or satellite images of the particular field. The additional data can also be used to generate an uncovered incorporation for one or more agricultural fields through a convolution neural network.
[0181] In step 810, one or more predicted yield values for the one or more agricultural fields are computed from the identification of the particular plantation from the effect of identifying the plantation on plantation yield, the environmental effect on planting yield and the effect of plantation yield management practice using a master neural network configured using plantation identification effects on plantation yield, environmental effects on plantation yield and effects of plantation yield management as input and plantation yield data as an exit. For example, a master neural network can be configured to accept incorporations for one or more agricultural fields as inputs and produce one or more yield values as outputs.
[0182] 4.2. RECOMMENDATIONS [0183] In one embodiment, the deep neural network 700 is used to generate recommendations for a particular field. For example, a user of a field manager computing device can request seed recommendations for a particular field through an application
Petition 870190123722, of 11/26/2019, p. 92/105
86/90 running on the field manager computing device. The user can indicate a location of the particular field.
[0184] In response to the request, the server computer system can identify one or more entries for the particular field. For example, the server computer system can access one or more yield maps passed to the field. As another example, the server computer system may request ground maps for the field from the field manager computing device or from one or more other computing devices. The server computer can also generate estimates for one or more entries. For example, the server computer system may request weather forecasts and / or meteorological measurements passed to the particular field from one or more external sources, such as a field manager computing device, one or more sensors, or an external server computer. Based on meteorological information, the server computer can generate temporal environmental data for the particular field.
[0185] After the server computer system has established one or more entries as being static, the server computer system can run the neural network with different entries in order to identify higher throughput values. For example, the server computer system can run the neural network with genotypes from different cultures and different management practices in order to identify a recommendation for the particular field. In one embodiment, the server computer system can store data that
Petition 870190123722, of 11/26/2019, p. 93/105
87/90 identify a range of ideal management practices for different types of seeds and / or different weather conditions. Using the stored data, the server computer system can minimize the number of runs of the deep neural network 700 by using only management practices that have been identified as ideal for seed and / or weather patterns.
[0186] The server computer system can identify the best seed to plant based on continuous runs of the deep neural network 700. The type of seed that produces total yield, risk-adjusted yield, planting quality values and / or profits Higher totals can be selected as the recommended crop. In addition, if each type of seed were introduced into the neural network with different management practices, the method described in this document would allow the recommendation of a management practice for the selected type of seed.
[0187] Although the example set out above represented the use of the neural network to recommend types of seeds, the neural network can also be used to recommend management practices for a particular area. For example, the user of the field manager computing device can identify the seed to be planted. The server computer system can identify or estimate values for environmental data through external sources. The server computer system can also identify yield maps passed to the particular field in order to improve recommendation accuracy. Using the fixed inputs, the
Petition 870190123722, of 11/26/2019, p. 94/105
88/90
computer server can perform the neural network with Appetizer many different for data of practices in management. 0 system server computer can to recommend the data in practice of management what
resulted in the highest requested yield value.
[0188] 4.3. NEW INFORMATION-BASED PREDICTIONS [0189] An additional quality of the deep neural network 700 is the ability to generate predictions based on new values for inputs that were not initially used to generate the neural network. As an example, the deep neural network 700 can be used to generate predictions as to how a new type of seed would react to different types of soil, different types of weather conditions and different management practices. The predictions would allow optimization in the use of new types of seeds or generation of new types of seeds without performance of physical tests.
[0190] In one embodiment, the genome for the new seed type is used as an input to the deep neural network 700. As the deep neural network 700 has been trained in several genomes, the deep neural network 700 is configured to be able to accept new types of genomes as inputs. The server computer system can then run the neural network with static values for environmental data and data from management practices to determine how the new seed type would perform under such conditions.
[0191] In one embodiment, the deep neural network 700 can also be used to identify ideal or less than ideal conditions for the new type of seed. For example,
Petition 870190123722, of 11/26/2019, p. 95/105
89/90 the deep neural network 700 can identify soil types that work best for a particular seed. In order to identify the best soil types for the new seed, the server computer system can select a plurality of soil types for testing. For each type of soil, the server computer system can run the deep neural network 700 a plurality of times with different inputs for weather conditions and management practices. The server computer system can identify which soil types consistently produced the highest yield values and which soil types consistently produced the lowest yield values. The server computer system can perform similar methods to determine the best and worst management practices and conditions for the seed.
[0192] In one embodiment, the deep neural network 700 can be used to determine the susceptibility of the new seed to changes in weather conditions, soil conditions or management practices. Similar to the identification of ideal conditions as described above, the server computer system can select small changes in a condition type and execute the deep neural network 700 repeatedly for each small change. If yield values change considerably from one temperature value to the next temperature value, then the seed can be identified as being highly sensitive to changes in temperature. Alternatively, if yield values change very little across a temperature range, then the seed can be
Petition 870190123722, of 11/26/2019, p. 96/105
90/90 identified as less sensitive to changes in temperature.
[0193] 5. BENEFITS OF CERTAIN MODALITIES [0194] Using the techniques described in this document, a computer system can compute yield value based on a large amount of data of various types in a computationally efficient way. Specifically, different neural networks allow the creation of a master neural network that can accept different types of data input. The incorporations described in this document reduce the computational cost of training the master neural network by reducing the number of inputs to the master neural network. Additionally, the use of a neural network makes it possible to reduce the processing capacity required of the computing device in relation to that which would be required for a comprehensive model of agricultural data.
[0195] 6. EXTENSIONS AND ALTERNATIVES [0196] In the specification described above, modalities were described with reference to numerous specific details that may vary from implementation to implementation. The specification and drawings, therefore, are to be considered in an illustrative rather than a restrictive sense. The unique and exclusive indicator of the scope of the disclosure, and what is proposed by the claimants to be the scope of the disclosure, is the literal and equivalent scope of the set of claims that result from this application, in the specific form in which such claims result, including any subsequent correction.

权利要求:
Claims (16)
[1]
1. Method, characterized by the fact that it comprises:
receiving, in a server computer system, a set of private data relating to one or more agricultural fields, in which the private data set comprises identification data of particular plantations, particular environmental data and data of particular management practices;
using a first neural network configured using plantation identification data as input and plantation yield data as output, compute a plantation identification effect on plantation yield for the one or more agricultural fields from the identification data of particular plantations ;
using a second configured neural network that uses environmental data as input and plantation yield data as an output, to compute an environmental effect on plantation yield for the one or more agricultural fields from the particular environmental data;
using a third configured neural network that uses management practice data as input and plantation yield data as output, to compute a management practice effect on plantation yield for the one or more agricultural fields from the management practice data private individuals;
using a configured master neural network that uses plantation identification effects on plantation yield, environmental effects on plantation yield, and management practice effects on yield
Petition 870190071515, of 7/26/2019, p. 14/22
[2]
2. Method, according to claim 1, characterized by the fact that:
the identification data for particular plantations comprises one or more genome sequences for one or more plantations corresponding to the particular data set;
the first neural network comprises a recurrent neural network configured to identify parts of the genome sequences that are correlated with effects on crop yield data.
2/8 plantation as input and plantation yield data as output, compute one or more predicted yield values for the one or more agricultural fields from the plantation identification effect on plantation yield, from the environmental effect on plantation yield and the effect of management practice on plantation yield.
[3]
3/8 the particular environmental data comprises one or more time series of predicted meteorological events and one or more spatial maps of soil properties;
the second neural network comprises a recurrent neural network for meteorological events and a convolution neural network for soil properties;
computing the environmental effect on plantation yield comprises generating an environmental incorporation learned using the recurrent neural network for meteorological events and the convolution neural network for soil properties.
3. Method, according to claim 2, characterized by the fact that the recurrent neural network is a neural network of long-term short-term memory.
[4]
4/8 private satellite in plantation yield for the one or more agricultural fields.
4. Method, according to claim 2, characterized by the fact that the recurrent neural network is a neural network of recurrent gate units.
[5]
5/8 plantation using environmental data as input;
a third neural network stored in memory, configured to compute a management practice effect on plantation yield using data from management practices as input;
a master neural network stored in memory, configured to compute one or more yield values using the plantation identification effect on plantation yield, the environmental effect on plantation yield and the effect of management practice on plantation yield as inputs;
one or more processors communicatively coupled to memory, configured to execute one or more instructions to cause performance of:
receive a private data set relating to one or more agricultural fields, where the private data set comprises identification data for particular plantations, private environmental data and private management practice data;
using the first neural network, compute a particular plantation identification effect on plantation yield for the one or more agricultural fields from the identification data of particular plantations;
using the second neural network, to compute a particular environmental effect on plantation yield for the one or more agricultural fields from the particular environmental data;
using the third neural network, compute a particular management practice effect on
Petition 870190071515, of 7/26/2019, p. 18/22
5. Method according to claim 1, characterized by the fact that the one or more predicted yield values comprise one or more of a risk adjusted yield value, a total profit value or a plantation quality value.
[6]
6/8 planting for one or more agricultural fields from the data of particular management practices;
using the master neural network, compute one or more predicted yield values for the one or more agricultural fields from the effect of identifying a particular plantation on plantation yield, the particular environmental effect on plantation yield and the effect of management practice plantation yield.
6. Method, according to claim 1, characterized by the fact that:
Petition 870190071515, of 7/26/2019, p. 15/22
[7]
7/8
7. Method, according to claim 1, characterized by the fact that:
the particular data set additionally comprises one or more particular satellite images of one or more agricultural fields at particular periods of a growing season;
the method further comprising:
using a fourth neural network configured using satellite images of fields at particular periods of the growing season as input and plantation yield data as output, to compute a particular satellite image effect on plantation yield for the one or
more agricultural fields a leave one or more images in private satellites; where the network neural master is configured additionally to use It is made of image in satellite in
plantation yield as input;
where the one or more predicted yield values are computed additionally from the image effect of
Petition 870190071515, of 7/26/2019, p. 16/22
[8]
8/8 in which the master neural network is additionally configured to use the satellite image effect in plantation yield as an input;
wherein the one or more predicted yield values are computed additionally from the particular satellite image effect on plantation yield for the one or more agricultural fields.
8. Method, according to claim 1, characterized by the fact that it additionally comprises:
the particular data set further comprising one or more maps of particular past yields of the one or more agricultural fields;
the method further comprising:
using a fourth neural network configured using past yield maps as input and plantation yield data as output, to compute a particular past yield effect on plantation yield for the one or more agricultural fields from the one or more past yield maps private individuals;
wherein the master neural network is additionally configured to use past yield effect on plantation yield as an input;
wherein the one or more predicted yield values are computed additionally from the particular past yield effect on plantation yield for the one or more agricultural fields.
[9]
9. System, characterized by the fact that it comprises:
a memory;
a first neural network stored in memory, configured to compute a plantation identification effect on plantation yield using plantation identification data as input;
a second neural network stored in memory, configured to compute an environmental effect on performance
Petition 870190071515, of 7/26/2019, p. 17/22
[10]
10. System, according to claim 9, characterized by the fact that:
the identification data for particular plantations comprises one or more genome sequences for one or more plantations corresponding to the particular data set;
the first neural network comprises a recurrent neural network configured to identify parts of the genome sequences that are correlated with effects on crop yield data.
[11]
11. System, according to claim 10, characterized by the fact that the recurrent neural network is a neural network of long-term short-term memory.
[12]
12. System according to claim 10, characterized by the fact that the recurrent neural network is a neural network of recurrent gate units.
[13]
13. System, according with The claim 9, characterized by fact that the one or more values in predicted yields understand one or more of a value in adjusted income risk, a value in profit total or one
plantation quality value.
Petition 870190071515, of 7/26/2019, p. 19/22
[14]
14. System, according to claim 9, characterized by the fact that:
the particular environmental data comprises one or more time series of predicted meteorological events and one or more spatial maps of soil properties;
the second neural network comprises a recurrent neural network for meteorological events and a convolution neural network for soil properties;
the second neural network is additionally configured to compute the environmental effect on plantation yield by generating an environmental incorporation learned using the recurrent neural network for meteorological events and the convolution neural network for soil properties.
[15]
15. System, according to claim 9, characterized by the fact that it additionally comprises:
a fourth neural network stored in memory, configured to compute a satellite effect on crop yield using satellite images as input;
wherein the particular data set further comprises one or more particular satellite images of one or more agricultural fields at particular periods of a growing season;
where the one or more processors are additionally configured to execute one or more instructions to cause performance of:
using the fourth neural network, to compute a particular satellite image effect on plantation yield for the one or more agricultural fields from one or more particular satellite images;
Petition 870190071515, of 7/26/2019, p. 20/22
[16]
16. System according to claim 9, characterized by the fact that it additionally comprises:
a fourth neural network stored in memory, configured to compute a past yield map effect on plantation yield using past yield maps as input;
wherein the particular data set further comprises one or more particular past performance maps of one or more agricultural fields;
where the one or more processors are additionally configured to execute one or more instructions to cause performance of:
using a fourth neural network, to compute a particular past yield effect on plantation yield for the one or more agricultural fields from the one or more particular past yield maps;
wherein the master neural network is additionally configured to use past yield effect on plantation yield as an input;
wherein the one or more predicted yield values are computed additionally from the particular past yield effect on plantation yield for the one or more agricultural fields.

类似技术:

---

公开号 | 公开日 | 专利标题

---

BR112019015401A2|2020-03-31|PLANTATION YIELD ESTIMATE USING AGRONOMIC NEURAL NETWORK

---

US10827669B2|2020-11-10|Method for recommending seeding rate for corn seed using seed type and sowing row width

---

US10859732B2|2020-12-08|Long-range temperature forecasting

---

US10990716B2|2021-04-27|Location selection for model assessment sampling

---

BR112019002763A2|2019-08-20|automatic detection of discrepancy values in collected data

---

US10684612B2|2020-06-16|Agricultural management recommendations based on blended model

---

BR112020022324A2|2021-01-26|digital visualization of seasonal agricultural fertility prescriptions updated periodically

---

BR112021006110A2|2021-07-20|systems and methods for identifying and using test sites in agricultural fields

---

US20210307243A1|2021-10-07|Cross-grower study and field targeting

---

BR112021001667A2|2021-05-04|automatic yield forecast and sowing rate recommendation based on weather data

---

BR112019018057B1|2021-10-26|COMPUTER IMPLEMENTED METHOD TO SELECT LOCATIONS IN A FIELD FOR TREATMENT SAMPLING

---

BR112019018015B1|2021-12-14|COMPUTER IMPLEMENTED METHOD TO SELECT SAMPLING LOCATIONS IN A FIELD AND IMPROVE GROWTH IN THE FIELD

---

US20200201269A1|2020-06-25|Utilizing spatial statistical models for implementing agronomic trials

---

BR112021002525A2|2021-05-04|enhancement of digital nutrient models using spatially distributed values unique to an agronomic field

同族专利:

---

公开号 | 公开日

---

US20180211156A1|2018-07-26|

---

EP3574467A1|2019-12-04|

---

EP3574467A4|2020-10-28|

---

AR110850A1|2019-05-08|

---

WO2018140225A1|2018-08-02|

---

US10699185B2|2020-06-30|

---

US20200334518A1|2020-10-22|

---

CA3051358A1|2018-08-02|

引用文献:

---

公开号 | 申请日 | 公开日 | 申请人 | 专利标题

---

---

US7702597B2|2004-04-20|2010-04-20|George Mason Intellectual Properties, Inc.|Crop yield prediction using piecewise linear regression with a break point and weather and agricultural parameters|

---

GB2464677A|2008-10-20|2010-04-28|Univ Nottingham Trent|A method of analysing data by using an artificial neural network to identify relationships between the data and one or more conditions.|

---

US9140824B1|2015-01-23|2015-09-22|Iteris, Inc.|Diagnosis and prediction of in-field dry-down of a mature small grain, coarse grain, or oilseed crop using field-level analysis and forecasting of weather conditions, crop characteristics, and observations and user input of harvest condition states|

---

US10514799B2|2016-09-08|2019-12-24|Google Llc|Deep machine learning to perform touch motion prediction|

---

EP3340130A1|2016-12-23|2018-06-27|Hexagon Technology Center GmbH|Method for prediction of soil and/or plant condition|WO2018049289A1|2016-09-09|2018-03-15|Cibo Technologies, Inc.|Systems for adjusting agronomic inputs using remote sensing, and related apparatus and methods|

---

EP3582603A4|2017-03-21|2021-01-06|Blue River Technology Inc.|Combine harvester including machine feedback control|

---

US9984455B1|2017-06-05|2018-05-29|Hana Resources, Inc.|Organism growth prediction system using drone-captured images|

---

US10580093B2|2017-06-17|2020-03-03|Institute Of Agricultural Resources And Regional Planning Chinese Academy Of Agricultural Sciences|Estimating the prior for a stochastic allocation model|

---

US10477756B1|2018-01-17|2019-11-19|Cibo Technologies, Inc.|Correcting agronomic data from multiple passes through a farmable region|

---

CN108986470A|2018-08-20|2018-12-11|华南理工大学|The Travel Time Estimation Method of particle swarm algorithm optimization LSTM neural network|

---

CN109034496B|2018-08-29|2020-11-06|中国水利水电科学研究院|Slope furrow optimal parameter determination method for retarding slope agricultural non-point source pollution|

---

US20200125929A1|2018-10-19|2020-04-23|X Development Llc|Crop yield prediction at field-level and pixel-level|

---

WO2020084414A1|2018-10-22|2020-04-30|Radient Technologies Innovations Inc.|Yield and market analytics|

---

US20200134486A1|2018-10-24|2020-04-30|The Climate Corporation|Leveraging genetics and feature engineering to boost placement predictability for seed product selection and recommendation by field|

---

US11240961B2|2018-10-26|2022-02-08|Deere & Company|Controlling a harvesting machine based on a geo-spatial representation indicating where the harvesting machine is likely to reach capacity|

---

US11178818B2|2018-10-26|2021-11-23|Deere & Company|Harvesting machine control system with fill level processing based on yield data|

---

CN109558939B|2018-10-29|2021-02-12|广东奥博信息产业股份有限公司|Crop growth model selection method and device based on neural network|

---

US10990875B2|2018-11-28|2021-04-27|International Business Machines Corporation|Neural network forecasting for tiered hydroponic natural farming configurations|

---

US11120552B2|2019-02-27|2021-09-14|International Business Machines Corporation|Crop grading via deep learning|

---

US11079725B2|2019-04-10|2021-08-03|Deere & Company|Machine control using real-time model|

---

US11234366B2|2019-04-10|2022-02-01|Deere & Company|Image selection for machine control|

---

AU2020312596A1|2019-07-17|2021-12-02|Farmers Edge Inc.|Automatic crop classification system and method|

---

WO2021067257A1|2019-09-30|2021-04-08|Nutrien Ag Solutions, Inc.|Agriculture service platform|

---

CN110728446A|2019-10-09|2020-01-24|中国地质大学（武汉）|County scale crop yield estimation method based on CNN-LSTM|

---

EP3816879A1|2019-11-04|2021-05-05|Gaf AG|A method of yield estimation for arable crops and grasslands and a system for performing the method|

---

US20210183108A1|2019-12-16|2021-06-17|X Development Llc|Edge-based crop yield prediction|

---

US11243332B2|2020-06-24|2022-02-08|X Development Llc|Predicting climate conditions based on teleconnections|

---

CN111990052B|2020-08-24|2021-09-03|农业农村部南京农业机械化研究所|Grain impurity-containing and breakage rate detection device of rotary disc type combine harvester and granary|

---

CN113435649A|2021-06-29|2021-09-24|布瑞克农业大数据科技集团有限公司|Global agricultural data sorting method, system, device and medium|

法律状态:
2021-10-13| B350| Update of information on the portal [chapter 15.35 patent gazette]|

优先权:

---

申请号 | 申请日 | 专利标题

---

US15/416,694|US10699185B2|2017-01-26|2017-01-26|Crop yield estimation using agronomic neural network|

---

US15/416,694|2017-01-26|

---

PCT/US2018/012949|WO2018140225A1|2017-01-26|2018-01-09|Crop yield estimation using agronomic neural network|

---