• 专利附录
  • 专利说明
  • 权利要求
  • 类似技术
  • 同族专利
  • 引用文献
  • 法律状态
  • 优先权
    • 专利摘要:
    it is a weed control system (2) for an agricultural sprayer (1) comprising a camera (3) and a spray unit (4) with several supply modules, a nozzle (9) and a module controller to receive a weed species detection signal and control the spraying of the chemical agent. the weed control system (2) also comprises a weed species identification unit (5) with a communication module, a memory module and a processing module that has several parallel processing cores. each parallel processing core performs a convolution operation between a submatrix constructed from pixels close to the image and a predefined core stored in the memory module to obtain a characteristic submatrix of the image's pixel values. the processing module calculates the probability of the presence of a weed species from the characteristic representation matrix and generates a weed species detection signal.
    公开号:BR112019016111A2
    申请号:R112019016111
    申请日:2018-02-06
    公开日:2020-04-14
    发明作者:Jourdain Guillaume；Serrat Hugo；Beguerie Jules
    申请人:Bilberry Sas；
    IPC主号:
    专利说明:

    “SYSTEMS AND METHODS FOR THE CONTROL OF WEED AND AGRICULTURAL SPRAYER THAT INCORPORATES THE SAME”
    FIELD OF INVENTION [0001] The present invention relates to weed control systems for agriculture and farming, in particular agricultural sprayers, and methods for applying pesticides in agriculture using such weed control systems and methods for calibrate these weed control systems.
    BACKGROUND OF THE INVENTION [0002] In particular, the present invention relates to weed control systems for agriculture and crops, in particular agricultural sprayers.
    [0003] Chemical methods for weed control often involve the application of a weed control agent, such as a herbicide or bio-herbicide. Pesticides from agricultural crops can be applied before emergence or after emergence with regard to the germination state of the plant and help to reduce the competitive pressure on newly germinated plants by removing unwanted organisms and maximizing the amount of water, soil nutrients and sunlight available for the crop.
    [0004] In agriculture, large-scale and systematic procedures for the application of pesticides are generally necessary and performed by large equipment known as tractor mounted or trailed sprayers.
    [0005] A sprayer generally comprises at least one spear with nozzles at intervals along the spear, a tank to contain the mixture of water and chemicals and a pump to distribute the herbicide to the nozzles through tubes. The sprayers come in various types, self-propelled, towed by a tractor, mounted on a tractor or demountable (mounted on a tractor tool holder, for example).
    Petition 870190074696, of 8/2/2019, p. 104/148
    2/32 [0006] To reduce the environmental impact as well as the cost of the weed control operation, there have been recent developments in vision-guided tractor that are provided with optical sensors, for example, color images to guide spray equipment bandwidth. The spraying system of tractors or mechanical cultivation devices are then controlled to treat only a weed when it is detected in the soil.
    [0007] The commonly used guidance technology takes advantage of the known pattern in which a crop was planted to distinguish between crops and weeds. Color images are, for example, transformed into grayscale images so that the green plant looks shiny against a background of dark soil. The plant / row spacing is entered into the computer and a predefined planting pattern can be combined with the grayscale image during tractor movement. The randomly distributed weed is then identified as a green area outside the regular crop pattern.
    [0008] However, such systems have several drawbacks.
    [0009] First, the crop pattern is often irregular or may evolve over the crop field in such a way that a predefined plant / row spacing inserted at the beginning of the operation becomes irrelevant. There may also be very few crop rows (for example, if there is a high rate of crop failure) or too many weeds in an image to reliably match the crop pattern. Such systems are also not able to use different selective herbicides for different weed species.
    [0010] There are attempts to combine such systems with analysis of the spectrum of the weed emission light or format recognition algorithms that compare the shape of the observed weed with formats stored in a database of weeds.
    Petition 870190074696, of 8/2/2019, p. 105/148
    3/32 [0011] Such systems have several disadvantages that prevent their use in practice.
    [0012] First, while the weed emission light spectrum can be characterized under laboratory conditions, the spectral reflectance of plants varies greatly under natural conditions, depending on weed species and soil variations. Naturally occurring weeds have a great diversity of species that can differ greatly from industrially grown weeds. The composition of the soil also has a strong effect on the color of the weeds and their spectral reflectance. As a result, the weed emission light spectrum cannot generally be used as a reliable criterion for distinguishing between weed species.
    [0013] Secondly, given the great diversity of weed species, a database of weed shapes has to be very large to store all variations of different shapes (young weed, old weed shapes, of all subspecies ...). The format recognition algorithms that compare the weed shape observed with the formats stored in the weed shape database are therefore very slow, as they need to scan these large databases.
    [0014] Third, the analysis of the weed light emission spectrum requires a level of brightness that makes it difficult to use such systems when the brightness is low, especially at dawn, dusk or even at night, when it would be desirable apply certain types of treatments. The low level of brightness and the speed of the spray induce a noise problem when detecting the green color, which has a negative impact on the sharpness of the acquired images and the processes that can be applied based on the acquired images.
    [0015] As a consequence, sprayers can travel only at a very low speed to allow sufficient time to
    Petition 870190074696, of 8/2/2019, p. 106/148
    4/32 the weed identification algorithm to process the images acquired by the camera and identify the weed species.
    [0016] WO2016025848A1 discloses a control system for agriculture and farming, specifically designed to perform a support count of crops such as corn.
    [0017] This reference sometimes mentions the detection of weeds, but does not teach any approach to the detection of weed species or the identification of weeds. In addition, the means of detection disclosed in this document are based on the spectral response of the plant in the visible or near infrared range.
    [0018] In addition, document WO2013059399A1, cited in the reference above, describes merely additional details about the spectral detection mentioned above.
    [0019] There is, therefore, a need for a weed control system capable of identifying weed species at a higher speed and with greater precision and allowing real-time spray control for selective herb treatment, even in difficult conditions, such as high speed of spraying and lighting, and even to identify small plants located in the most remote part of the images (upper part of the image with an inclined axis of view).
    SUMMARY OF THE INVENTION [0020] For this purpose, according to the invention, this weed control system for an agricultural sprayer comprises:
    [0021] * at least one camera adapted to be mounted on an agricultural sprayer (1) to acquire an image of a portion of a crop field, while said agricultural sprayer travels in a crop field, said image comprises an array of pixel values,
    Petition 870190074696, of 8/2/2019, p. 107/148
    5/32 [0022] * a spray unit adapted to be mounted on said agricultural sprayer and comprising [0023] - at least one supply module comprising a chemical agent tank, [0024] - at least one spray nozzle a chemical agent from said at least one supply module and [0025] - a controller module adapted to receive a weed species detection signal and to selectively control the spraying of the chemical agent from said supply module via of at least one nozzle based on said weed species detection signal, [0026] and the weed control system is characterized by additionally comprising:
    [0027] * a weed species identification unit [0028] - a communication module adapted to receive the image acquired by the camera and send a weed species detection signal to a controller module of a sprayer unit weed control system, [0029] - a memory module adapted to store said image, and [0030] - a processing module comprising a plurality of parallel processing cores, [0031] with each processing core Parallel is adapted to perform at least one convolution operation between a submatrix constructed from pixels close to the image and a predefined core of reference pixel images stored in the memory module to obtain a submatrix representing the characteristics of the pixel values of the image, [0032] and the processing module is
    Petition 870190074696, of 8/2/2019, p. 108/148
    6/32 adapted to calculate at least one probability of the presence of a weed species among a database of weed species from an image characteristic representation matrix constructed from the characteristics representation sub-matrices constructed by parallel processing cores; and to generate a weed species detection signal based on said at least one probability of presence.
    [0033] Thanks to the invention, several characteristics can be taken into account at the same time in the recognition of weed species, such as, for example, the shape, texture, color and / or the location of the weed species in the images , thanks to an implementation of artificial intelligence processing involving cores.
    [0034] Other optional and non-limiting features of the invention include the following:
    [0035] * a target area of the spraying unit and a field of view of the camera are separated from each other by a separation distance d s along a longitudinal axis of the agricultural sprayer, [0036] where the number of processing cores of the weed species identification unit is such that a weed species detection signal can be generated from an image acquired by the camera with a latency time value of such that:
    ds / (tl + tp)> V [0037] where v is a speed value of the agricultural sprayer moving in the crop field and P is a time value of the spray unit.
    [0038] * the camera has a longitudinal extension of the FOVx field of view along a longitudinal geometric axis of the selected agricultural sprayer so that a signal of detection of species of
    Petition 870190074696, of 8/2/2019, p. 109/148
    7/32 weeds can be generated from an image acquired by the camera with a ti latency time, such that:
    FOVx / ti> v [0039] where v is the speed of the agricultural sprayer (1) traveling in the crop field.
    [0040] * the definition of the camera and the position of the camera in the agricultural sprayer are selected in such a way that each pixel of an image generated by said camera independently covers an elementary area of the soil area of less than five square millimeters.
    [0041] * each parallel processing core performs said at least one convolution operation by calculating a matrix for matrix multiplication between the submatrix and a predefined core matrix, or by calculating a fast Fourier transform of the submatrix, being that the parallel processing cores perform rapid Fourier convolutions or transforms based on a given image from the camera at the same time.
    [0042] * the parallel processing cores of the processing module are adapted to perform a grouping operation of the image representation matrix comprising the determination of a statistics of a submatrix of values close to the characteristics representation matrix, in particular a maximum value of said submatrices of close values.
    [0043] The parallel processing cores of the processing module are adapted to perform a non-linear parallel processing of the image representation matrix by applying a non-linear operation to each value of said representation matrix, such as a rectified linear activation function.
    [0044] * the processing module is adapted to perform a sequence of n processing operations starting with a
    Petition 870190074696, of 8/2/2019, p. 110/148
    8/32 input matrix generated from the pixel values of the image and ending with an output matrix that comprises at least one probability of the presence of a weed species among a database of weed species, in particular, wherein each processing operation / of said sequence of successive processing operations takes as input a representation matrix F1-i emitted by a previous processing operation i-1 or an input matrix generated from the pixel values of the image, and generates a representation matrix of F characteristics.
    [0045] * the weed species detection signal comprises an weed species location indicator, [0046] in particular, where the output matrix comprises at least one value indicating a weed species location weed among the database of weed species within the image acquired by the camera.
    [0047] * a location of weed species determined from the weed species location indicator is stored in the memory module.
    [0048] * the spray unit comprises a plurality of nozzles arranged along a transverse direction of the agricultural sprayer, and [0049] in which the spray unit controller module is adapted to receive the weed species detection signal and to further control the spraying of chemical agent through the plurality of nozzles based on the weed species location indicator.
    [0050] * the spraying unit comprises at least one movable nozzle adapted to selectively spray a plurality of target zones along a transverse direction of the agricultural sprayer, and
    Petition 870190074696, of 8/2/2019, p. 111/148
    9/32 [0051] in which the spray unit controller module is adapted to receive the weed species detection signal and to additionally control a position and / or orientation of the mobile nozzle based on the weed species location indicator weeds.
    [0052] The present invention additionally provides an agricultural sprayer comprising a weed control system, as defined above, mounted on said agricultural sprayer.
    [0053] According to another aspect, the present invention provides a method for the control of weeds using an agricultural sprayer as defined above which comprises:
    [0054] acquire an image of a portion of a crop field, while said agricultural sprayer is traveling in a crop field, using at least one camera mounted on the agricultural sprayer, said image comprising a matrix of pixel values, [0055] receiving the image acquired by the camera in a communication module of a weed species identification unit of the agricultural sprayer and storing said image in a memory module of said herb species identification unit weeds, [0056] run in parallel, in a plurality of respective cores of parallel processing of a processing module, a plurality of respective convolution operations, being that each convolution operation is executed between a submatrix built from pixels close to the image and a predefined core stored in the memory module to obtain a representation submatrix of characteristics of the pixel values of the image, [0057] calculate at least one probability of the presence of a weed species among a species database
    Petition 870190074696, of 8/2/2019, p. 112/148
    10/32 of weeds from a matrix of representation of characteristics of the image constructed from the submatrices of representation of characteristics constructed by the cores of parallel processing, [0058] generate a signal of detection of weed species based on the said at least a probability of the presence of a weed species and send said weed species detection signal to a controller module of a weed control system spray unit, [0059] receiving the weed detection signal weed species in a controller module of a spray unit mounted on the agricultural sprayer, and [0060] selectively control the spraying of the chemical agent from at least one spray unit supply module through at least one nozzle based on the weeds species detection signal.
    [0061] Finally, the present invention provides a method for calibrating a weed control system as defined above and adapted to spray a plurality of weed species listed in a weed species database, where [0062 ] a vehicle is provided with at least one camera adapted to acquire an image of a portion of a culture field, during a movement of said vehicle in a culture field, comprising the said image comprising a matrix of pixel values , [0063] said vehicle moves in a crop field that has at least a predefined number of each weed species from a target weed species database and acquires at least a predefined number of images of each species weed from said target weed species database,
    Petition 870190074696, of 8/2/2019, p. 113/148
    11/32 [0064] a set of training data is constructed from said predefined number of images of each weed species, marking said weed species in said images, [0065] a set of weights model weed identification determined from the training data set, said set comprising at least one predefined core for a convolution operation performed by a parallel processing core of a weed control system as defined above, [ 0066] where the weed identification model weight set is stored in a weed control system memory module, as defined above.
    BRIEF DESCRIPTION OF THE DRAWINGS [0067] Other characteristics, objectives and advantages of the invention will appear easily from the following description of several of its modalities, provided as non-limiting examples, and the accompanying drawings.
    [0068] In the drawings:
    [0069] - Figure 1 is a schematic perspective view of an agricultural sprayer comprising a weed control system according to an embodiment of the invention, [0070] - Figure 2 is a block diagram illustrating the modules and the weed control system units in Figure 1, [0071] - Figures 3 and 4 are two flowcharts that detail the processing operations performed by the processing module of a weed species identification unit in the weed control of Figures 1 and 2, and [0072] - Figure 5 is a flowchart that details a step in a method for calibrating the weed control system of
    Petition 870190074696, of 8/2/2019, p. 114/148
    12/32
    Figures 1 and 2 according to the modalities of the invention.
    [0073] In the different Figures, the same reference signs designate the same or similar elements.
    DETAILED DESCRIPTION [0074] Figure 1 illustrates an example of agricultural sprayer 1 according to the invention.
    [0075] Agricultural sprayer 1 is designed to move in a crop field. In particular, the agricultural sprayer 1 can be mounted or dragged by a tractor. The sprayer can be self-propelled, can be towed by a tractor, mounted on a tractor or demountable (also equipped with a conveyor, for example).
    [0076] A culture field is illustrated in Figure 1 and generally features C cultures that can be organized along rows and can be visible or not (pre-emergent cultures, for example). A variety of W weeds can also grow in the field, as illustrated in Figure 1.
    [0077] The agricultural sprayer 1 is designed to move, along a longitudinal X direction in the crop field, usually along one or more rows of crops in the field. The agricultural sprayer 1 comprises at least one boom 17 which extends along a transverse direction Y perpendicular to the longitudinal direction X.
    [0078] The agricultural sprayer 1 is provided with a weed control system 2 according to the invention, which is illustrated schematically in Figure 2.
    [0079] More precisely, the weed control system 2 comprises one or more cameras 3, a spray unit 4 and a weed species identification unit 5.
    [0080] Cameras 3 are mounted on agricultural sprayer 1, for example, on the agricultural sprayer boom, and are adapted to acquire images of a portion of the crop field while the sprayer
    Petition 870190074696, of 8/2/2019, p. 115/148
    13/32 agricultural 1 moves in the field. The camera can be mounted with a certain angle in relation to a vertical direction perpendicular to the transverse and longitudinal direction, for example, about 45 or 60 degrees, in order to acquire an image of the crop field in front of the agricultural sprayer. Cameras 3 can be mounted on dampers in order to reduce vibrations during the movements of the sprayer 1 and increase the sharpness of the acquired images. As will be explained below, several characteristics can be taken into account in the recognition of weed species, including the shape, texture, color and / or location of the weed species in the images. The sharpness of the acquired images is also important to obtain this information, esp. although the light is low and the speed of the sprayer is fast. Several criteria may therefore be necessary to recognize a wide variety of weed species, and the present invention allows for this.
    [0081] Each camera 3 comprises a sensor 6, such as a CCD or CMOS sensor, and an optical system 7 comprising a plurality of lenses and / or mirrors.
    [0082] A camera 3 acquires an image that comprises an array of pixel values. Each image comprises W * H pixels, where W is the number of pixels over an image width and H is a number of pixels over an image height. The W width and H height of the camera define a camera sensor resolution. The sensor may, in particular, acquire an image comprising an array of at least one million pixel values, more preferably, more than 5 million pixel values.
    [0083] Alternatively, a linear camera can be used and an array of pixel values can be reconstructed from the output of the linear camera.
    [0084] Advantageously, the location of the camera, the resolution of the sensor and the design of the optical system are selected so
    Petition 870190074696, of 8/2/2019, p. 116/148
    14/32 that the image acquired by the camera comprises a matrix of at least one million pixels, in which each pixel of said matrix independently covers an area of the ground of less than five square millimeters, preferably less than two square millimeters. This spatial resolution of the camera is important to enable the reliable identification of weed species, as detailed below.
    [0085] Camera 3 can be a color camera. In this case, the pixel values include, for example, three channels, such as RGB values (red-green-blue) and the pixel value matrix is a 3D matrix of dimensions W * H * 3, for example.
    [0086] In contrast to the previously known weed control systems, camera 3 can be free of color filters and polarizer. In particular, the optical system 7 of the camera 3 can consist only of lenses and / or mirrors. Since only the general color and no absolute spectral information is needed to identify the weed, a bare camera can be used and provided only with selected lenses and mirrors to obtain the spatial resolution necessary for the identification of weed species.
    [0087] A spray unit 4 is also mounted on agricultural sprayer 1 and comprises several components which will now be described in greater detail.
    [0088] As shown in Figure 2, the spray unit 4 comprises at least one supply module 8 with at least one nozzle 9 and a controller module 10. In particular, the spray unit 4 can comprise at least two supply modules 8 that contain different chemicals.
    [0089] Each supply module 8 comprises a chemical agent tank 11 and an electronic distribution valve 12.
    [0090] The chemical agent tank 11 contains a liquid that can be a herbicide or water.
    Petition 870190074696, of 8/2/2019, p. 117/148
    15/32 [0091] The tank may contain a premixture of water and herbicide or a separate tank of water may contain water to be mixed with a herbicide during operation, during or immediately before delivery. Each supply module 8 can be supplied with a tank 11 containing different liquid. For example, a supply module 8 can be supplied with a tank 11 containing a herbicide containing an active agent adapted for weed treatment before weed germination, while another supply module 8 can be supplied with a tank 11 which contains a herbicide with active agents adapted for weed treatment after the emergence of weeds.
    [0092] The spray unit 4 comprises at least one nozzle 9 for spraying a chemical agent from at least one of said supply modules 8 in a target zone S of the field.
    [0093] The nozzles are mounted on the agricultural sprayer, for example, spread along the transverse direction of the agricultural sprayer extension, as shown in Figure 1.
    [0094] The spraying unit controller module 10 receives a weed species detection signal from the weed species identification unit 5, as will be described below. Based on this signal, the controller module 10 selectively controls the spraying of the chemical agent from at least one of the supply modules 8 through the nozzle 9.
    [0095] The controller module 10 can control the spraying of the chemical agent with a spray delay after receiving the weed species detection signal.
    [0096] The spray delay can be calculated based on the latency time of the weed recognition unit, as well as the speed of the vehicle 1 and the calibrated distance between the nozzle 9 and the camera 3.
    Petition 870190074696, of 8/2/2019, p. 118/148
    16/32 [0097] The spray delay can also take into account a pre-calibrated delivery time of the chemical agent corresponding to the latency of the mechanical systems of the spray unit and the time of displacement of the liquid in the spray unit tubes, for example example.
    [0098] In one embodiment of the invention, the weed species detection signal only comprises information regarding the weed species. The controller module 10 of the spray unit 4 then selects a supply module 8 that contains a herbicide adapted for this species of weeds and controls the electronic distribution valve 12 and, if necessary, the nozzle 9, to spray the agent chemical.
    [0099] In other embodiments of the invention, the weed species detection signal may comprise an indicator of weed species location.
    [0100] In this modality, a location of weed species in the field can be determined from said weed species location indicator, for example, using additional information about the location of sprayer 1 obtained using a global positioning system and, optionally, additional calibration information on the orientation and / or position of camera 3 on sprayer 1.
    [0101] The location of the weed species in the field can then be stored in a memory, in particular, in memory module 14.
    [0102] In one of those embodiments illustrated in the Figures, the spraying unit 4 comprises a plurality of nozzles 9 arranged along the transverse direction Y of extension of the agricultural sprayer. Nozzles 9 are directed towards the field and each nozzle 9 is adapted to spray a chemical to cover a target zone S of the field.
    [0103] In this mode, controller module 10 of
    Petition 870190074696, of 8/2/2019, p. 119/148
    17/32 spray unit 4 can then receive the weed species detection signal and control the spraying of the chemical agent through the plurality of nozzles 9 based on the weed species location indicator. In particular, only a limited number of nozzles 9 can be opened according to the location of the weed in the soil.
    [0104] In a variant, the spray unit 4 can comprise at least one movable nozzle 9. The movable nozzle can be adapted to selectively spray a plurality of target zones S on the ground, for example, a plurality of target zones S juxtaposed along the transversal direction Y of the agricultural sprayer 1. The mobile nozzle 9 can have its position and / or orientation controlled by a tilting or sliding control unit.
    [0105] In this mode, the controller module 10 of the spraying unit 4 receives the weed species detection signal and controls the position and / or orientation of the mobile nozzle 9 based on the weed species location indicator. In particular, the position and / or orientation of the nozzle 9 can be adapted according to the location of the weed in the soil, in order to spray the herbicide in the correct location in the soil.
    [0106] The two modalities described above can be combined in order to provide a plurality of movable nozzles independently moved and selected.
    [0107] Reference is now made more specifically to Figure 2, which illustrates in more detail a weed species identification unit 5 according to an embodiment of the invention.
    [0108] The weed species identification unit 5 comprises a communication module 13, a memory module 14 and a processing module 15.
    [0109] Communication module 13 receives the image acquired by camera 3 and sends the detection signals for species of
    Petition 870190074696, of 8/2/2019, p. 120/148
    18/32 weeds for the controller module 10 of the spray unit 4 of the weed control system 2 as previously described.
    [0110] In particular, the camera 3 can generate a continuous flow of images during the movement of the agricultural sprayer 1 in the crop field, in which case the communication module 13 can continuously receive said images and all the modules of the identification unit Weed Species 5 can work in real time or smooth real time, also generating and sending a continuous stream of weed species detection signals to the controller module 10 of the spray unit 4.
    [0111] The communication module 13 can communicate with the camera 3 and with the controller module 10 of the spray unit 4 by wired communication or using a wireless protocol (for example, optical or radio protocol, such as infrared or “Wi-Fi”).
    [0112] The memory module 14 is capable of storing the received image or image stream. The memory module 14 can comprise several submodules and can be distributed on several chips of the weed species identification unit 5. In particular, the memory module 14 can comprise a non-volatile storage memory and a volatile storage memory.
    [0113] The processing module 15 comprises a plurality of parallel processing cores p 16. The number p of parallel processing cores 16 is greater than one. The processing module can, for example, comprise at least four parallel processing cores 16. The parallel processing cores 16 can perform parallel calculations on different submatrices and cores, as will be described in more detail below.
    [0114] Each parallel processing core 16 may comprise a plurality of sub-core processors in order to further parallelize image processing.
    Petition 870190074696, of 8/2/2019, p. 121/148
    19/32 [0115] In particular, parallel processing cores can be part of a single computing component 15, for example, a central processing unit (CPU) or a graphics processing unit (GPU).
    [0116] The parallel processing core 16 may have access to the specific area of the memory module 14, in particular, the memory module 14 may include a memory chip located next to the processing module 15, for example, a memory chip of a graphics processing unit that incorporates the processing module 15.
    [0117] A basic operation of a processing core 16 is a convolution operation between a given submatrix P constructed from pixels close to the image and a predefined core K stored in memory module 14, to obtain a submatrix F of representation of characteristics of the pixel values of the image. For example, the pixel dimensions of each pixel submatrix close to the image are selected to be identical to the dimensions of the predefined stored core K.
    [0118] The K core is a small matrix that can be considered as an equivalent receptive field for a given pixel in the image.
    [0119] The convolution operation involves the calculation of the dot product between the K-core inputs and a submatrix P constructed from pixels close to the image to produce a characteristic representation submatrix, which is a filtered representation of the image.
    [0120] As explained earlier, a submatrix P of the image contains pixels close to that image. Each image can, therefore, be divided into several submatrices P, so that the scalar product calculated between the nucleus K and each submatrix P can, typically, generate characteristic representation submatrices. It is then possible to build a representation matrix of characteristics from these sub-matrices of
    Petition 870190074696, of 8/2/2019, p. 122/148
    20/32 representation of characteristics.
    [0121] Given the location of the convolution operation, convolutions can be easily parallelized in a multi-core hardware environment, which greatly speeds up image processing.
    [0122] In addition, as detailed below, the coefficients of a core can be calibrated so that the core is general and can be applied to a wide variety of weed species. Determining the core coefficients is part of determining the weights of the weed identification model. This means that the convolution operation performed with the K core must allow the recognition of each weed species. The core coefficients are therefore representative of the characteristics that characterize each weed species.
    [0123] To distinguish between different weed species, several characteristics can be considered in isolation or in combination, such as the shape, texture, color and / or the location of the weed species in the images. The core of coefficients must therefore be calibrated according to these parameters or characteristics. Taking into account a sufficient number of characteristics that distinguish the different species of weeds, the effectiveness of the recognition of weed species and the speed of said recognition are improved. This is particularly advantageous to allow the sprayer to move faster and reduce the total treatment time.
    [0124] In particular, to perform said convolution operation, each parallel processing core can calculate a matrix-to-matrix multiplication between the submatrix P and a predefined core matrix K associated with the core.
    [0125] The coefficients of the K-matrix matrix can be identical throughout the image and, therefore, identical between the nuclei of
    Petition 870190074696, of 8/2/2019, p. 123/148
    21/32 parallel processing or may vary depending on the location of the submatrix processed in the image.
    [0126] This matrix for matrix multiplication can be parallelized using conventional parallel processing algebra algorithms to increase the image processing speed.
    [0127] Alternatively, the convolution operation can be performed by calculating a fast Fourier transform of the submatrix P of the image.
    [0128] As illustrated in Figures 3 and 4, the processing module generally performs a sequence of n processing operations starting with an input matrix I generated from the pixel values of the image and ending with an output matrix What comprises at least a probability of the presence of a weed among a database of weed species.
    [0129] Advantageously, each processing operation / of said sequence of successive processing operations takes as input a characteristic representation matrix Fi-i emitted by a previous processing operation i-1 or the input matrix I generated from the pixel values of the image and generates a representation matrix of F characteristics.
    [0130] Processing operations involve at least one of the following:
    [0131] - a convolution operation as described previously, [0132] - a grouping operation and / or [0133] - non-linear parallel processing.
    [0134] The grouping operation and the nonlinear parallel processing will now be described in further detail.
    [0135] A grouping operation can be
    Petition 870190074696, of 8/2/2019, p. 124/148
    22/32 performed by each parallel processing core of the processing module.
    [0136] A grouping operation can be performed on a submatrix S of close values determined from the input matrix I or a characteristic representation matrix F, -i issued by a previous processing operation i-1.
    [0137] A characteristic representation matrix Fi can be constructed from the characteristic representation submatrices obtained by applying a convolution operation between the K core and P submatrices of the input matrix I or the characteristic representation matrix Fm. Likewise, the matrix of representation of characteristics Fi can be divided into several sub-matrices S of close values. Likewise, the input matrix I can be divided into several sub-matrices S of close values.
    [0138] A grouping operation can be applied to each submatrix S of close values. For example, it is possible to obtain sub-matrices of characteristic representation as sub-matrices S of close values when the grouping operation is applied to a characteristic representation matrix.
    [0139] The grouping operation is a local operation to reduce the size of a representation matrix of characteristic Fi-ι or of the input matrix I, preserving the most important information. For example, for each submatrix S of close values, only one value is retained. In other words, after applying the grouping operation, a reduced characteristic representation matrix Fi is obtained so that the characteristic representation matrix Fi contains only, for example, a value for each submatrix S of the representation matrix characteristic Fm or input matrix I.
    [0140] The grouping operation involves the determination of a statistic of said submatrix S of close values. THE
    Petition 870190074696, of 8/2/2019, p. 125/148
    23/32 statistics is, for example, a maximum value of said submatrix S, as in the so-called “maximum grouping”. In the mode in which a maximum grouping operation is used, only the maximum value of each submatrix S is retained.
    [0141] Since the grouping operation is a local operation, it can also be easily parallelized and increases the robustness of the weed species identification in relation to a small change of the weed in the image between the formation images and the images of test.
    [0142] The parallel processing cores of the processing module are also capable of performing a non-linear parallel processing of the input matrix I or a representation matrix F1-i emitted by a previous processing operation i-1.
    [0143] By "non-linear operation", it is understood that the output y = f (x) of the non-linear function / applied to a scalar, vector or tensor x is not linear in relation to said scalar, vector or tensor x.
    [0144] An example of a non-linear function is a rectified linear unit, such as the function f (x) = max (0, x) or a generalized rectified linear unit, such as a leaking rectified linear unit, a parametric rectified linear unit or a maxout unit. For example, the generalized function can be:
    f (x) = max (0, x) + a * min (0, x) [0145] where a is a predefined parameter.
    [0146] The non-linear function can be applied independently to each value of the input matrix I or the characteristic representation matrix F-i.
    [0147] Unlike the grouping operation, the non-linear operation can preserve the size of the input matrix I or the characteristic representation matrix Fi-i.
    [0148] Here, again, the independent application of
    Petition 870190074696, of 8/2/2019, p. 126/148
    24/32 non-linear function at each value of the input matrix I or the characteristic representation matrix F -1 makes processing easily parallelized and thus reduces the latency of the weed identification unit.
    [0149] The successive processing operation can thus lead to an output matrix containing probabilities of detecting each weed species in the weed species database.
    [0150] In some embodiments of the invention, the output matrix may also comprise at least one value indicative of a location of a weed species within the image acquired by the camera.
    [0151] This allows you to select and / or move the nozzle to reduce the consumption of chemicals.
    [0152] Advantageously, all identified weed species can be provided with location information.
    [0153] Such a value indicating a location of a weed species can, for example, be a bounding box indicating a location of the weed species within the image acquired by the camera.
    [0154] From the output matrix, processing module 15 is thus capable of calculating at least one probability of the presence of a weed species among the weed species database.
    [0155] The processing module 15 can thus generate a weed species detection signal based on the said probability of presence.
    [0156] Using the special operation and parallel processing described above, it is thus possible to obtain a weed identification system with very low latency.
    Petition 870190074696, of 8/2/2019, p. 127/148
    25/32 [0157] More precisely, weed species identification unit 5 can be adapted to generate a weed species detection signal from an image acquired by camera 3 with a ti latency time.
    [0158] The latency time ti corresponds to a time that separates the generation of the weeds species detection signal from the reception of the corresponding image I.
    [0159] The latency time ti can be less than 500 ms, in particular, less than 200 ms or even less than 100 ms, with a corresponding detection accuracy greater than 75%, in particular, greater than 90%.
    [0160] Detection accuracy means the number of detection among weed species observed in a large number of images, for example, more than 1,000 images (that is, the number of true positives over the total number of weed specimens that appear in said images).
    [0161] In addition, a target zone S of the spray unit 4, in particular of the nozzle 9 of the spray unit, and a FOV field of view of the camera 3 can be separated from each other by a separation distance ds to the along the longitudinal axis X of displacement.
    [0162] The separation distance ds and the latency time ti can be such that ds / (ti + tp)> v, where v is a speed of the agricultural sprayer 1 traveling over the crop field and P is a spray unit processing time.
    [0163] The spray unit processing time t P is the time between receipt at the spray unit of the information that a weed has been identified and the actual spraying of the herbicide. This time may, in particular, comprise the delay detailed above. The processing time of the spray unit t p is, for example,
    Petition 870190074696, of 8/2/2019, p. 128/148
    26/32 on the order of 200 ms or less.
    [0164] In one mode, the agricultural sprayer can travel in the crop field with a speed v, for example, between 7 and 25 km / h. The target zone S of the spray unit and the FOV field of view of the camera can be located very close to each other along the longitudinal direction of the path, for example, between 1 and 6 m away. The latency time ti can therefore be around 200 ms, for example.
    [0165] The latency time ti can also be restricted by camera 3, as will be described now. The camera 3 acquisition system has a predefined field of view (FOV). More precisely, the focal length of the lens and the size of the image sensor establish a relationship between the field of view and the working distance (the distance between the back of the lens and the imaged part of the culture field). The FOV field of view is therefore the inspection area captured by the camera's sensor. The size of the field of view and the size of the camera's sensor directly affect image resolution (a determining factor in accuracy). The field of view is, in particular, limited by the resolution required to enable the identification of weed species, as detailed above.
    [0166] The FOV field of view can therefore be expressed in square meters and can, in particular, be less than 10 square meters, for example, around 5 square meters.
    [0167] The field of view extends along the longitudinal direction X and the transverse direction Y.
    [0168] The longitudinal extent of the FOVx field of view can be between 1 m and 3 m. The transversal extension of the FOVy field of view can be between 1 m and 5 m.
    [0169] Latency time can also be limited by the longitudinal extent of the FOVx field of view and the frame rate of the camera.
    Petition 870190074696, of 8/2/2019, p. 129/148
    27/32 [0170] For example, if the camera has a longitudinal extension of the FOVx field of view of about 1 meter, the camera must produce a new image whenever the agricultural sprayer has traveled 1 meter along the longitudinal direction. In order to avoid image accumulation in temporary storage, the latency time of the weed species identification unit must be less than the time between two consecutive camera acquisitions 3.
    [0171] In particular, a relationship can be defined between the longitudinal extension of the field of view (FOVx) and the latency time (ti) as follows:
    FOV X / lt x > v [0172] or equivalent:
    t < F0V * / <tv [0173] For example, if the longitudinal extension of the field of view (FOVx) is about 1 meter and the speed (v) is about 20 km / h, that is, 5.5 m / s, the latency time (ti) must be less than 180 ms.
    [0174] The parameters of the processing operations described above, in particular the parameters of the cores of the convolution operations, can be determined by operating a calibration process that will now be described in more detail.
    [0175] A plurality of weed species is listed in a weed species database. The weed species database may comprise, for example, Cirsium arvense Scop, Chenopodium polispermum L, Bromus sterilis L., Papaver rhoeas L., Datura stramonium L, Avena fatua L., Galium aparine L., Geranium dissectum L , Sonchus oleraceus L Convolvulus arvensis L., Matricaria sp., Polygonum convolvulus L, Veronica hederaefolia L., Alopecurus agrestis L.
    [0176] A vehicle, such as agricultural sprayer 1, is equipped with at least one camera 3 adapted to acquire an image of
    Petition 870190074696, of 8/2/2019, p. 130/148
    28/32 a portion of a culture field, during movement of said vehicle in the culture field.
    [0177] Camera 3 used during the calibration process can be similar to the camera described above.
    [0178] Vehicle 1 is designed to move in a crop field. It can be similar to the agricultural sprayer described above. The vehicle can be supplied only with a camera and therefore without a spray unit 4 or a weed species identification unit 5.
    [0179] Camera 3 acquires images that comprise arrays of pixel values as detailed above.
    [0180] Vehicle 1 travels in a crop field that has at least a predefined number of each weed species from a target weed species database. The predefined number is advantageously a large number, for example, greater than a few hundred or thousands of samples of each target weed species.
    [0181] Vehicle 1 thus acquires at least a predefined number of images of each weed species in said target weed species database. For example, more than a few hundred or a few thousand images that contain a sample of each target weed species. This step of acquiring at least a predefined number of images of weed species is the first step in the process illustrated in Figure 5.
    [0182] In a second step, a set of training data can be constructed from the predefined number of images of each weed species by marking the weed species in the images. The marking operation can include the assignment of a category of weeds to each sample acquired in the image and can also include the definition of a bounding box or an indication of
    Petition 870190074696, of 8/2/2019, p. 131/148
    29/32 location, within an image, of each weed species shown in said image.
    [0183] For each category of weed, the system can therefore use several samples that illustrate the said category of weeds. It is then possible to determine common characteristics among said samples, such as the shape, texture, color and / or location of the category of weeds. A learning step described below is based on the training data set and the marked weed species to which a category has been assigned.
    [0184] In other words, from the images in which the different weed species have been indicated, it is possible to determine the distinguishing characteristics that will allow the different weed species to be distinguished from each other. Therefore, it is not even necessary to indicate on which criteria the weed species will be differentiated, since these criteria can be determined automatically by analyzing images and determining common characteristics between several samples marked as indicating the same weed species.
    [0185] In a third step, a set of weights identification model weights is then determined from the training data set. The weed identification model weight set comprises at least one predefined core for a convolution operation, as detailed above.
    [0186] In fact, the core coefficients need to be calibrated so that the core is general and can be applied to a wide variety of weed species. The core coefficients are determined based on the training data set. Again, the core coefficients are determined based on the characteristics of the different weed species learned based on the training data set, such as the shape, texture, color and / or location of the weeds.
    Petition 870190074696, of 8/2/2019, p. 132/148
    30/32 species of weeds in the images.
    [0187] The images acquired by vehicle 1, also called training data set, allow to learn the characteristics of the weed species to determine a set of weights identification model weights, as well as the core coefficients. This learning step is performed to maximize the accuracy of the model. The objective of this step is, for example, to maximize the probability of predicting the weed samples marked in the training on the data set. The set of model weights can be determined using machine learning techniques, for example, using descending gradient algorithms. The operations described above are performed on the images of the training data set. The core coefficients used initially to perform the operations can be determined in different ways. For example, the core coefficients can be predefined at random. It is then possible to determine an error rate on the images on which the operations were performed. In fact, as the different weed species were marked in the images, it is possible to compare, for each identified weed species, if the detection obtained by performing the operations is correct. If the error rate is not acceptable, for example, if the error rate is greater than a predetermined limit, a reverse propagation learning can be carried out to modify the weights identification model weight set, hence the coefficients of the core. Obviously, after the first pass, significant modifications to the weights of the weed identification model are necessary, especially if these parameters have been preset at random. This step can obviously be repeated as many times as necessary.
    [0188] To summarize the above, the training data set is used for a learning stage during which the weights of operations and the core coefficients are determined. Features such as the shape, texture, color and / or location of each
    Petition 870190074696, of 8/2/2019, p. 133/148
    31/32 weed species are determined automatically based on the images in the training data set where a marking operation was performed to assign a category of weeds to each sample. After performing operations on the images in the training data set, the accuracy of the model is estimated, with an error rate, for example, and backpropagation learning is performed to modify the weights of the weed identification model. This step of performing operations and reverse propagation learning can be repeated so that the error rate obtained is reduced.
    [0189] Finally, in a fourth step, the set of weights from the weed identification model is stored in memory module 14 of weed control system 2 and can then be used for a weed treatment operation as detailed above.
    [0190] As will be well understood by those skilled in the art, the various steps and processes discussed here to describe the invention may refer to operations performed by a computer, processor or other electronic calculation device that manipulates and / or transforms data with use electrical phenomenon. These computers and electronic devices may employ a variety of volatile and / or non-volatile memories, including non-transient, computer-readable media, with an executable program stored on it including various executable codes or instructions for execution by the computer or processor, where the memory and / or the computer-readable medium may include all forms and types of memory and other computer-readable media.
    [0191] The previous discussion disclosed and describes merely exemplary modalities of the present invention. One skilled in the art will readily recognize from such a discussion and the attached drawings and claims that various changes, modifications and variations can be made without departing from the spirit and scope of the invention as defined in the
    Petition 870190074696, of 8/2/2019, p. 134/148
    32/32 claims to follow.
    权利要求:
    Claims (15)
    [1]
    1. Weed control system (2) for an agricultural sprayer (1) comprising:
    * at least one camera (3) adapted to be mounted on an agricultural sprayer (1) to acquire an image of a portion of a crop field while said agricultural sprayer travels in a crop field, said image comprising a matrix of pixel values, * a spray unit (4) adapted to be mounted on said agricultural sprayer and comprising
    - at least one supply module (8) comprising a chemical agent tank (11),
    - at least one nozzle (9) for spraying a chemical agent from said at least one supply module (8), and
    - a controller module (10) adapted to receive a weed species detection signal and to selectively control the spraying of the chemical agent from said supply module through at least one nozzle based on said species detection signal of weeds, and the weed control system (2) is characterized by the fact that it additionally comprises:
    * a weed species identification unit (5) comprising
    - a communication module (13) adapted to receive the image acquired by the camera (3) and to send a weed species detection signal to a controller module (10) of a spray unit (4) of the control system of weeds,
    - a memory module (14) adapted to store said image and
    - a processing module (15) comprising a
    Petition 870190074696, of 8/2/2019, p. 136/148
    [2]
    2/7 plurality of parallel processing cores (16), with each parallel processing core (16) being adapted to perform at least one convolution operation between a submatrix built from pixels close to the image and a predefined image core of reference pixels stored in the memory module (14) to obtain a submatrix of characteristic representation of the pixel values of the image, and the processing module (15) is adapted to calculate at least one probability of the presence of a kind of weeds among a database of weed species from a characteristic representation matrix of the image constructed from the characteristic representation submatrices constructed by the parallel processing cores; and to generate a weed species detection signal based on said at least one probability of presence.
    2. Weed control system according to claim 1, characterized by the fact that a target zone (S) of the spray unit (4) and a field of view (FOV) of the camera (3) are separate from each other by a separation distance d s along a longitudinal geometric axis (X) of the agricultural spraying device (1), where the number of processing cores of the weed species identification unit (5) is such that a weed species detection signal can be generated from an image acquired by the camera (3) with a latency time value ti such that: ds / (tl + tp)> v where v is a speed value of the agricultural sprayer (1) traveling in the crop field et P is a processing time value of the spray unit (4).
    [3]
    3. Weed control system according to claim 2, characterized by the fact that the camera (3) has a longitudinal extension of the FOVx field of view along a geometric axis
    Petition 870190074696, of 8/2/2019, p. 137/148
    3/7 longitudinal (X) of the agricultural sprayer (1) selected so that a weed species detection signal can be generated from an image acquired by the camera (3) with a ti latency time, such that:
    FOVx / ti> v where v is the speed of the agricultural sprayer (1) traveling in the crop field.
    [4]
    4. Weed control system according to any one of claims 1 to 3, characterized in that the camera definition (3) and the position of the camera in the agricultural sprayer (1) are selected so that each pixel of an image generated by said camera independently cover an elementary area of the ground area of less than five square millimeters.
    [5]
    5. Weed control system according to any one of claims 1 to 4, characterized by the fact that each parallel processing core (16) performs said at least one convolution operation by calculating a matrix for multiplication of matrix between the submatrix and a predefined core matrix, or by calculating a fast Fourier transform of the submatrix, with the parallel processing cores performing rapid convolutions or Fourier transforms based on a given image from the camera at the same time.
    [6]
    6. Weed control system according to any one of claims 1 to 5, characterized in that the parallel processing cores (16) of the processing module (15) are adapted to perform a matrix grouping operation of representation of characteristics of the image that comprises the determination of a statistic of a submatrix of values close to said matrix of representation of characteristics, in particular, a maximum value of said submatrix of close values.
    [7]
    7. Weed control system according to any one of claims 1 to 6, characterized by the fact that the nuclei
    Petition 870190074696, of 8/2/2019, p. 138/148
    4/7 parallel processing (16) of the processing module (15) are adapted to perform a nonlinear parallel processing of the characteristics representation matrix of the image by applying a non-linear operation to each value of said characteristics representation matrix , such as a rectified linear activation function.
    [8]
    8. Weed control system according to any one of claims 1 to 7, characterized in that the processing module (15) is adapted to carry out a sequence of n processing operations starting with a generated input matrix from the pixel values of the image and ending with an output matrix that comprises at least one probability of the presence of a weed species among a database of weed species, in particular, where each processing / of said sequence of successive processing operations takes as input a representation matrix of characteristics Fi-i emitted by a previous processing operation i-1 or an input matrix generated from the pixel values of the image, and generates a matrix of representation of Fi characteristics.
    [9]
    9. Weed control system according to any one of claims 1 to 8, characterized by the fact that the weed species detection signal comprises an indicator of weed species location, in particular, in which the output matrix comprises at least one value indicating a location of a weed species among the database of weed species within the image acquired by the camera.
    [10]
    10. Weed control system, according to claim 9, characterized by the fact that a location of weed species determined from the weed species location indicator is stored in the memory module (14).
    Petition 870190074696, of 8/2/2019, p. 139/148
    5/7
    [11]
    Weed control system according to claim 9 or 10, characterized in that the spraying unit (4) comprises a plurality of nozzles (9) arranged along a transverse direction (Y) of the sprayer (1), and where the controller module (10) of the spray unit (4) is adapted to receive the weed species detection signal and to additionally control the spraying of chemical agent through the plurality of nozzles (9 ) based on the weed species location indicator.
    [12]
    Weed control system according to any one of claims 9 to 11, characterized in that the spraying unit (4) comprises at least one mobile nozzle (9) adapted to selectively spray a plurality of zones- target along a transverse direction (Y) of the agricultural sprayer (1), and in which the controller module (10) of the spray unit (4) is adapted to receive the weed species detection signal and to additionally control a position and / or orientation of the mobile nozzle (9) based on the weed species location indicator.
    [13]
    13. Agricultural sprayer (1) characterized by the fact that it comprises a weed control system (2) according to any one of claims 1 to 12, mounted on said agricultural sprayer.
    [14]
    14. Method for controlling weeds using an agricultural sprayer (1), according to claim 13, characterized by the fact that it comprises acquiring an image of a portion of a crop field while said agricultural sprayer ( 1) is moving in a crop field, using at least one camera (3) mounted on the agricultural sprayer, the said image comprising a matrix of pixel values,
    Petition 870190074696, of 8/2/2019, p. 140/148
    6/7 receive the image acquired by the camera (3) in a communication module (13) of a weed species identification unit (5) of the agricultural sprayer and store said image in a memory module (14) of the said weed species identification unit, carry out in parallel, in a plurality of respective parallel processing cores (16) of a processing module (15), a plurality of respective convolution operations, with each convolution operation being performed between a submatrix built from pixels close to the image and a predefined core stored in the memory module (14) to obtain a submatrix representing the characteristics of the pixel values of the image, calculate at least a probability of the presence of a species of grass weed among a database of weed species from a matrix of representation of characteristics of the constr image uted from the submatrices of representation of characteristics constructed by the parallel processing nuclei, generate a weed species detection signal based on said at least a probability of the presence of a weed species and send said weed detection signal weed species for a weed control system spraying unit control module, receiving weed species detection signal on a spraying unit (4) controlling module (10) mounted on the agricultural sprayer (1), and selectively control the spraying of the chemical agent from at least one supply module (8) of the spray unit (4) through at least one nozzle (9), based on the species detection signal of the weeds.
    [15]
    15. Method for calibrating an herb control system
    Petition 870190074696, of 8/2/2019, p. 141/148
    7/7 weeds (2) according to any one of claims 1 to 12, adapted to spray a plurality of weed species listed in a weed species database characterized by the fact that a vehicle is equipped with at least one camera (3) adapted to acquire an image of a portion of a culture field, during a movement of said vehicle (1) in a culture field, said image comprising a matrix of pixel values, the said vehicle travels in a crop field that has at least a predefined number of each weed species in a target weed species database and acquires at least a predefined number of images of each weed species in said database of target weed species, a set of training data is built from the predefined number of images of each weed species, marking said species of and weeds in said images, a set of weights identification model weights is determined from the training data set, said set comprising at least one predefined core for a convolution operation performed by a processing core parallel of a weed control system according to any one of claims 1 to 12.
    the weed identification model weight set is stored in a weed control system memory module (2) according to any one of claims 1 to 12.
    类似技术:
    公开号 | 公开日 | 专利标题
    BR112019016111A2|2020-04-14|weed control systems and methods and agricultural sprayer that incorporates the same
    US10761211B2|2020-09-01|Plant treatment based on morphological and physiological measurements
    Khan et al.2018|Estimation of vegetation indices for high-throughput phenotyping of wheat using aerial imaging
    EP2736318B1|2016-08-17|Plant growth kinetics captured by motion tracking
    Devia et al.2019|High-throughput biomass estimation in rice crops using UAV multispectral imagery
    JP2021516060A|2021-07-01|Plant cultivation data measurement method, work route planning method and equipment, system
    BR112019015015A2|2020-04-28|adaptive cyber-physical system for efficient monitoring of unstructured environments
    BR112020008086A2|2020-11-03|method for creating digital cropping map, digital cropping map, computer system and computer program product
    US20200011019A1|2020-01-09|Weeding systems and methods, railway weeding vehicles
    Lopes et al.2016|Vineyard yeld estimation by VINBOT robot-preliminary results with the white variety Viosinho
    US20210183108A1|2021-06-17|Edge-based crop yield prediction
    Nielsen et al.2012|Orchard and tree mapping and description using stereo vision and lidar
    Ponnusamy et al.2021|Precision agriculture using advanced technology of IoT, unmanned aerial vehicle, augmented reality, and machine learning
    Costa et al.2021|Measuring pecan nut growth utilizing machine vision and deep learning for the better understanding of the fruit growth curve
    Phillips2014|Precision agriculture: supporting global food security
    US20220075344A1|2022-03-10|A method of finding a target environment suitable for growth of a plant variety
    IL288358D0|2022-01-01|Automated plant monitoring systems and methods
    Roy2013|Above and below ground phenotyping methods for corn
    同族专利:
    公开号 | 公开日
    WO2018142371A1|2018-08-09|
    EP3357332A1|2018-08-08|
    EP3576526A1|2019-12-11|
    CA3050364A1|2018-08-09|
    AU2018215728A1|2019-08-08|
    US11071991B2|2021-07-27|
    US20200230633A1|2020-07-23|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    US4015366A|1975-04-11|1977-04-05|Advanced Decision Handling, Inc.|Highly automated agricultural production system|
    US5246164A|1991-12-16|1993-09-21|Mccann Ian R|Method and apparatus for variable application of irrigation water and chemicals|
    US5278423A|1992-12-30|1994-01-11|Schwartz Electro-Optics, Inc.|Object sensor and method for use in controlling an agricultural sprayer|
    DE19726917A1|1997-06-25|1999-01-07|Claas Selbstfahr Erntemasch|Device on agricultural machinery for contactless scanning of contours extending over the ground|
    US6785400B1|1999-08-17|2004-08-31|Image Therm Engineering, Inc.|Spray data acquisition system|
    DE10221948B4|2001-05-14|2004-03-11|Kümmel, Knut, Dipl., -Ing.|Process and system for volume-specific influencing of soil and plants|
    HU224536B1|2002-07-05|2005-10-28|G & G Növényvédelmi és Kereskedelmi Kft.|Connection arrangement of weeding equipment mounted onto a railway vehicle and weeding procedure|
    HU3056U|2005-04-29|2006-03-28|G & G Noevenyvedelmi Es Keresk|Construction for making weed map|
    US8437498B2|2010-08-31|2013-05-07|Valmont Industries Inc.|Self propelled mechanized irrigation system with a remote visual monitoring system|
    US8488874B2|2011-01-31|2013-07-16|Dalhouse University|Variable rate sprayer system and method of variably applying agrochemicals|
    WO2013059399A1|2011-10-20|2013-04-25|Monsanto Technology Llc|Plant stand counter|
    US20150245565A1|2014-02-20|2015-09-03|Bob Pilgrim|Device and Method for Applying Chemicals to Specific Locations on Plants|
    AU2015265625B2|2014-05-05|2020-10-08|Horticulture Innovation Australia Limited|Methods, systems, and devices relating to real-time object identification|
    WO2016025848A1|2014-08-15|2016-02-18|Monsanto Technology Llc|Apparatus and methods for in-field data collection and sampling|
    DE102015111889B4|2015-07-22|2020-02-27|Andreas Reichhardt|Procedure for testing an agricultural sprayer|
    US20170039425A1|2015-08-08|2017-02-09|Prospera Technologies, Ltd.|System and method for optimizing chemigation of crops|
    EP3455783A1|2016-05-12|2019-03-20|Basf Se|Recognition of weed in a natural environment|CN207266563U|2017-09-29|2018-04-24|好得科技（深圳）有限公司|A kind of Multifunction infrared senses water sprinkler|
    US10621434B2|2018-01-25|2020-04-14|International Business Machines Corporation|Identification and localization of anomalous crop health patterns|
    US11147258B2|2018-02-12|2021-10-19|Capstan Ag Systems, Inc.|Systems and methods for spraying an agricultural fluid on foliage|
    DE102018217742A1|2018-10-17|2020-04-23|Robert Bosch Gmbh|Process for applying a spray to a field|
    EP3667555A1|2018-12-14|2020-06-17|Bilberry SAS|A weeding system and method for use in agriculture|
    DE102019202041A1|2019-02-15|2020-08-20|Zf Friedrichshafen Ag|Method and agricultural vehicle for liquid application|
    EP3930453A1|2019-02-26|2022-01-05|Aventum GmbH|Device and method for removing or destroying weeds|
    FR3098684A1|2019-07-16|2021-01-22|Bilberry Sas|Plant treatment system in agriculture|
    WO2021090294A1|2019-11-09|2021-05-14|Bilberry Sas|Plant treatment system|
    WO2021211718A1|2020-04-17|2021-10-21|Bayer Cropscience Lp|Image monitoring for control of invasive grasses|
    DE102020205079A1|2020-04-22|2021-10-28|Robert Bosch Gesellschaft mit beschränkter Haftung|Method and device for applying at least one liquid medium by means of a spray nozzle unit and computer program product|
    GR1009982B|2020-04-26|2021-04-22|Παντελεημων Κωνσταντινου Κελεσης|Electronic spayer for linear cultivations|
    法律状态:
    2021-10-19| 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
    EP17305131.9A|EP3357332A1|2017-02-06|2017-02-06|Weed control systems and methods, agricultural sprayer|
    PCT/IB2018/050731|WO2018142371A1|2017-02-06|2018-02-06|Weed control systems and methods, and agricultural sprayer incorporating same|
    [返回顶部]