• 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:
    System and method of detection of anisakis parasites in fish fillets. The system comprises: - a lighting module (10) for subjecting the fish fillet (1) to pulsed lighting; - image capture means (20); - a control module (30) that synchronizes the capture of the image (3) with the pulsed lighting; - processing means (40) for: - apply a segmentation process on the captured image (3), obtaining a binary image (6); - perform a geometric identification of individuals present in the binary image (6) by means of correspondence of geometric patterns, obtaining geometric parameters; - extract intensity characteristics by spectral band of each individual, obtaining color parameters; - introduce the geometric parameters and coloration of each individual in an artificial neural network that establishes a relationship between the morphological, silhouette and color characteristics with the characteristics of anisakis, to estimate whether the individual is anisakis or not. (Machine-translation by Google Translate, not legally binding)
    公开号:ES2552405A1
    申请号:ES201430789
    申请日:2014-05-27
    公开日:2015-11-27
    发明作者:Joaquín GRACIA SALVADOR；Iñaki MINIÑO ARBILLA；Santiago PASCUAL DEL HIERRO；Ángel F. GONZÁLEZ GONZÁLEZ
    申请人:Luis Calvo Sanz Sa；Luis Calvo Sanz S A；Tecnologia Marina Ximo S L；TECNOLOGIA MARINA XIMO SL；
    IPC主号:
    专利说明:

    5 Field of the inventionAnisakis is a genus of parasitic nematodes, whose life cycle affects fish andmarine mammals, in which it can produce lesions in your digestive tract and pose atwo-way risk to human health: through infection by wormseat unprocessed fish and through allergic reactions to chemicals that
    10 worms leave in the fish.
    Anisakiasis frequently appears in areas of the world where fish is eaten raw or lightly salted or seasoned. Hours after ingestion of the parasite larvae, abdominal pain, nausea and vomiting may occur. If the larvae pass to
    15 intestine, a severe eosinophilic granulomatous response may occur even 1 02 weeks after infestation.
    The clinical manifestations of allergic reactions vary from urticaria and / or angioedema that are present in all of them, to cases of severe anaphylactic shock.
    The object of the invention is the identification of the anisakis prior to their intake by means of a compact system that allows to analyze samples of fish meat and cephalopods and identify the presence of the Anisakis parasite, determining the degree of infection.
    Background of the invention There are currently certain methods capable of detecting the presence of Anisakis in food. These methods are carried out in laboratories, using complex equipment designed for laboratory conditions that require large
    30 knowledge for handling, which are not isolated from water nor are they hygienic as they should be to work with food.
    Patent documents WQ2007118209-A2 and US20060129327A1 disclose pathogen detection procedures with complex operation and that present certain problems, in particular the need to chop the sample, centrifuge it,

    soak it in aqueous media or use certain plotters.
    The methods used are based on the identification of antigens, genetic and / or biochemical methods that require reagents and work in special laboratories that also make the fish used in the tests inedible. The present invention aims to eliminate the exposed problems, presenting a system and a method that allows to detect the presence of the Anisakis parasite in fish in an innocuous way, keeping the fish piece intact, so that the analyzed fish that is considered free of Anisakis It can be later ingested.
    Description of the invention The present invention relates to a high-resolution optical scanner-based equipment that, using optical, artificial vision and digital image processing techniques, is capable of specifically detecting the presence of Anisakis in fish samples and cephalopods analyzed.
    The identification of the anisakis is done prior to the intake of fish meat and cephalopods, determining the degree of infection and advising its disposal if necessary. In case of not having infection the same piece analyzed can be ingested since the analysis procedures are harmless.
    The analytical process begins with the acquisition of the image. This image is transferred to the internal processing module by means of algorithms that in addition to performing the previously described analysis, provides a tactile interface to the user to enter the information related to the test, the production lot, or any other significant information for the realization of a subsequent analysis or for the generation of reports.
    A first aspect of the present invention relates to an Anisakis parasite detection system in fish fillets, where the fish fillet to be inspected is deposited in an inspection area. The system includes:
    - a lighting module configured to subject the fish fillet to pulsed lighting; - image capture means configured to automatically capture a color image of the fish fillet while it is subjected to pulsed lighting; -a control module configured to synchronize the image capture with the
    pulsed lighting;-processing means configured for:
    • apply a segmentation process on the captured image, obtaining
    a binary image with the silhouette of different individuals that may correspond to 5 anisakis in the fish fillet;
    • make a geometric identification of each individual present in the binary image through a process of correspondence of geometric patterns, obtaining geometric parameters of each individual;
    • extract the characteristics of intensity and percentage of presence per spectral band 10 of each individual, obtaining color parameters of each individual;
    • introduce an input vector with the geometric and coloring parameters obtained for each individual in an artificial neural network that establishes a relationship between the morphological, silhouette and color characteristics of each individual with the characteristics of the anisakis, to estimate whether the individual is or
    15 not anisakis.
    In a preferred embodiment the lighting module is configured to generate the
    pulsed illumination within a wavelength band between 230 and 700 nanometers,
    with a controlled irradiance of up to 2 W / cm'l and / or in programmable periods with a
    20 ignition duration between 0.1 to 1 second per pulse and a shutdown duration between 0.1 to 5 seconds per pulse.
    The control module can be configured to synchronize the image capture with the illumination pulsed depending on the wavelength and irradiance
    25 programmed in the lighting module. To obtain the coloration parameters, the processing means can be configured to classify the intensity in different color bands between 230 and 700 nanometers of wavelength.
    30 The processing means are preferably configured to obtain the positions of the anisakis identified in the fish and their size. The lighting module may be formed by a plurality of high power LED emitters.
    For obtaining the binary image, the processing means are preferably configured to extract, from the captured image and by background segmentation techniques, the background belonging to the inspection area not hidden by the fish fillet, obtaining an image segmented fish fillet; and apply a thresholding process on said segmented image.
    5 The processing means may be configured to obtain the binary image
    by any one of the following ways:-a segmentation by wavelength band;-a segmentation by luminance;-a saturation segmentation;
    10-apply several of the previous segmentations and choice of segmentation that offers an image with the contours of the most defined individuals.
    The processing means can be configured to perform the geometric pattern matching process using a series of boundary curves that do not correspond to a pixel grid and looking for shapes similar to the patterns in the image without being based on specific gray levels.
    A second aspect of the present invention relates to a method of detecting Anisakis parasites in fish fillets, comprising:
    20-send the fish fillet to a pulsed illumination; - Automatically capture a color image of the fish fillet while it is subjected to pulsed lighting, the image being synchronized with the pulsed lighting;
    - apply a segmentation process on the captured image, obtaining a binary image with the silhouette of different individuals that may correspond to anisakis in the fish fillet;
    - make a geometric identification of each individual present in the binary image (6) through a process of correspondence of geometric patterns, obtaining geometric parameters of each individual;
    30-extract the characteristics of intensity and percentage of presence per spectral band of each individual, obtaining color parameters of each individual; -introduce an input vector with the geometric and coloration parameters obtained for each individual in an artificial neural network that establishes a relationship between the morphological, silhouette and color characteristics of each individual with the
    35 characteristics of anisakis, to estimate whether or not the individual is anisakis.
    The system presents the result of each process on the screen, storing both the analysis information and the information entered by the user in a database that allows subsequent consultation, printing and export to other formats (Excel, Word, etc.).
    The system has the following functionalities:
    • Acquisition of images from optical sensor.
    • Possibility of acquiring images from file (tif, jpg, bmp, etc ..).
    • Pre-processed for sample conditioning.
    • Automatic / manual selection on the image to be processed, of the tonal ranges of each of the elements to be detected and quantified.
    • Specific image processing algorithms for the identification, segmentation and quantification of Anisakis (complete individuals and fragments) against fish mass.
    • User interface, through a touch monitor, for the manual introduction of information related to the scanned species, weight, production lot, etc., necessary for further study or automatic report generation.
    • Indication by the user interface of the areas assigned to each of the elements:
    • Anisakis display on the image itself.
    • Indication of the number of Anisakis detected within the image.
    • Indication of the area and percentage of infection.
    • Indication of degree of infection in four states:
    1. Anisakis free.
    2. Mild infection
    3. Moderate infection
    Four. Intense infection.
    • Module for the automatic generation of reports for each of the samples processed by the system. This report may include the image with the masks identifying each of the specified and enhanced elements, and the information regarding the presence of each of them quantitatively (percentages, areas, no infection, mild, moderate and intense infection, etc. ...), together with the information entered by the user.
    • Storage of test results in a database.
    • Printing of customized reports (Iogo company, laboratory, configuration of report blocks, etc.).
    The present invention can be used, among others, for use in laboratories and professional purchase / sale control, for installation in processing lines of industrial plants, for use in fishmongers and other fish outlets, or for their use in restaurants and gourmet stores.
    BRIEF DESCRIPTION OF THE DRAWINGS A series of drawings that help to better understand the invention and that expressly relate to an embodiment of said invention that is presented as a non-limiting example thereof is described very briefly below.
    Figure 1 shows the elements of the detection system.
    Figure 2 shows an enlarged area of a captured image where anisakis can be seen in the fish.
    Figures 3A and 3B show, respectively, a captured image where the color band has been removed and the segmentation by wavelength band applied in said image.
    Figures 4A and 4B show, respectively, a captured image where the luminance and luminance segmentation applied in said image have been extracted.
    Figures 5A and 5B show, respectively, a captured image where saturation and saturation segmentation applied in said image have been extracted.
    DETAILED DESCRIPTION OF THE INVENTION The present invention consists of an automatic inspection system for the identification of Anisakis in fish fillets that integrates image acquisition, by means of an artificial vision system, and image processing, by means of algorithms supported by artificial neural networks.
    Figure 1 shows the different elements of the detection system. The systemcomprises a lighting module 10, image capture means 20 (a cameracolor image capture) and a control module 30 responsible for synchronizing theImage capture with lighting. The fish fillet 1 to be inspected is deposited in5 an inspection zone 2, for example a tray. The system comprises means ofprocessing 40 (which include for example a processor) that are responsible for analyzing theImages captured to detect anisakis. Such processing means maybe part of the control module 30 or be included in a separate unit. I also knowshown in Figure 1 a display module 50, which is optional, where
    10 show the images once analyzed by the processing means 40.
    The system is placed on a table or workbench. Fish fillets 1 are manually deposited in inspection tray 2. The system captures images of the scene. These images are stored and labeled according to the adjustment made
    15 by the operator through an interface / touch screen that incorporates the system.
    Once the image is captured, it is processed using a set of mathematical operations based on non-linear morphological and calorimetric models. Once the Anisakis are being detected and identified, a correlation model between size
    20 of the Anisakis in the image and its color intensity and wavelength is used to estimate the actual individual size. This process is carried out in parallel in all Anisakis detected in the fish fillet inspected.
    Each Anisakis is indicated in the image. The image is analyzed to extract two types of
    25 information: The positions or coordinates of the Anisakis in the fish fillet. The size of the Anisakis.
    These two individual parameters of each anisakis detected in a fish fillet are
    30 stored in a data file, with additional information, such as total fillet area, total Anisakis area, number of Anisakis present in the fillet, the species of fish analyzed, light and optical configuration parameters, time inspection and the identifier corresponding to the system that carried out the analysis.
    The hardware of the present invention is composed of a vision module (which in a particular embodiment consists of a high-resolution RGB industrial camera without a 2/3 inch sensor, working under the GigE Ethernet protocol), a lens fixed optics, a computer control system and all the electrical components necessary for safety and power.
    The system also has a lighting module, consisting of a set of special high-power LEO-based light source blocks, encapsulated in IP67 protection vessels with high cooling power. The LEO emitter design provides radiometric power greater than neon illumination in the specific wavelength range required for this application. Each emitter element has very small dimensions (about 4.4mm x 4.4mm), so this configuration offers an exceptional optical power density and minimizes stress, resulting in a highly reliable lighting module, very low maintenance and High radiant flow control. The objective of this module is to guarantee the most suitable lighting conditions to capture images. The structure that holds the light sources can be adjusted, in order to adapt to the dimensions of the inspection area, to ensure adequate and homogeneous illumination of the field of vision.
    The system is controlled by software developed specifically for this application. This software allows control of the camera parameter selection and allows the image registration program to work automatically. It also allows the interface other functionalities. The software implements the following functionality:
    Graphical user interface for inspection and control display (start or stop commands, image analysis display, cumulative inspection results, time and date, etc.)
    Graphical user interface for system configuration (work band setting, color model, threshold, camera diagnostic utility, inspection labeling, etc.). Configurable parameters:
    • Working band: Adjustable width from 230 to 700 nanometers. Parameter manually adjustable by combining optical filters in the light projectors and in the lens of the optical sensor.
    • Selection of the color model with which the optical sensor works:
    or HLS (Hue, Saturation, Lightness), which is an established color model, which combines the components of Hue, Saturation and Brightness.
    or RGB (Red, Green, Blue), which is a color model based on the
    5 additive synthesis, capable of obtaining a color by mixing by adding the three primary light colors, Red, Green and Blue.
    • Threshold adjustment: This is a numerical adjustment of the minimum amount of light signal needed by the optical sensor of the equipment in order to obtain an image of sufficient quality to be analyzed and
    10 processed by the anisakis detection algorithms. These parameters allow the equipment to be adjusted to obtain the best image capture so that the detection algorithms can work with the best and most defined information possible.
    15 A structured identification model in a model database, with all the parametric information necessary for the identification and estimation of individuals.
    A module for generating data analysis in the form of a file of
    20 text, with a header system identifier that includes an entry of each Anisakis detected and analyzed, with the following information: coordinate Anisakis position in the fish fillet, size of the Anisakis detected, total size of the fillet, total surface area of Anisakis detected , number of fillet parasitic anisakis, fish species, etc.
    25 A module for generating reports and wireless communication with other systems.
    For automatic image capture the fish fillet is subjected to lighting
    30 pressed in programmable periods of duration (TON) of 0.1 to 1 second, the programmable TOFF being 0.1 to 5 seconds with controlled emissions within a band that can range from 230 to 700 nanometers of wavelength and a controlled irradiance of up to 2 W / cm2, In the process of capturing the best image, lighting and its specific wavelength, power and pulse characteristics are essential
    35 so that the system can detect anisakis in fish fillets, not damaging them, which makes this invention a non-destructive anisakis detection system. Scanned fish fillets may deteriorate with some combinations of wavelength and irradiance if exposure is intense and prolonged. The high irradiance makes it possible to delve deeper into the fish fillet, so that the use of controlled pulses makes it possible to illuminate for short periods of time, sufficient for the optical sensor to capture a good image to process.
    An electronic system synchronizes the pulses of the lighting module with the captures of the optical sensor to capture at the most optimal time and thus obtain the best image that allows the process to detect the presence of anisakis in the fish fillet. The optimal moment of capture depends on the wavelength and irradiance programmed at each moment. The system automatically selects this moment and orders the optical sensor to capture.
    After automatic image capture and by background segmentation techniques, the bottom of the inspection tray that does not hide the fish fillet is extracted first and then on the segmented fillet the silhouette of each detected anisakis is extracted, to which it adjusts a sampling pattern to obtain precise indicators of the shape and contour of the individual or pieces thereof, typical of the anisakis. As there may be other parasites or elements that, in any case, have similar contour and shape, the processing of an artificial neural network is applied for the precise identification of anisakis.
    Neural networks are a combination of mathematical functions that simulate the way in which brain neurons connect to each other. In this way, an artificial neural network can be trained using iterative algorithms to model complex solution problems, establishing a relationship between a characteristic space (in this case all the information obtained from the inspection of the silhouette and color) with the different classes (the different forms detected in the fish fillet, within the inspection area).
    The information acquired by the optical sensors is derived to processing means responsible for detecting the elements that correspond to the fish.
    Next, algorithms developed exclusively for this application are applied. In the detection, the algorithms process the image and extract a series of correlation parameters, perform a pixel count to be able to extract the total area by anisakis and calculate distances to determine the length and width of each detected anisakis. Some of the basic correlation parameters extracted by the algorithms are, for each anisakis:
    • Total area.
    • Total length.
    • Average width.
    • Intensity detected band 700-600.
    • Intensity detected band 600-550.
    • Intensity detected band 550-450.
    • Intensity detected band 450-400.
    • Intensity detected band 400-230.
    • Percentage presence per spectrum band.
    Although the spectral bands used in a preferred embodiment are those indicated above (700-600: 600-550: 550-450: 450-400: 400-230), other spectral bands could be used (in numbers and limits, higher and lower).
    The invention is based on correlation of silhouettes parameters, estimation of volumes in multidimensional spaces of color representation from images with details at different wavelengths, and segmentation and filtering of captured images.
    Therefore, when identifying anisakis, a prior step is required to separate existing individuals from the already segmented fish area. This entails performing a new segmentation and dynamic identification. During this preprocessing phase the images are subjected to different subtraction and thresholding processes that involve various mathematical operations.
    Figure 2 shows, by way of example and in black and white, an enlarged area of a color captured image 3 obtained by the capture means of the inspection system of the present invention, where anisakis 4 can be seen in fish 1. In this image, bottom 5 (in black color), corresponding to inspection tray 2, is shown on the far left.
    Figure 3A shows a captured image (already without background 5) where the5 color band and Figure 3B shows the segmentation by wavelength bandapplied in the image of Figure 3A, obtaining a binary image 6.
    Figure 4A shows the same image captured from Figure 3A where the luminance has been extracted, and Figure 4B shows the luminance segmentation applied in image 10 of Figure 4A, obtaining a binary image 6.
    Figure 5B shows the saturation segmentation obtained from the captured image shown in Figure 5A, obtaining a binary image 6.
    15 In each segmentation calculations are made for the three processes indicated above (saturation, luminance and band) and the system determines which of the three techniques offers an image with the most appropriate and robust segmentation to segment the anisakis, analyzing the linearity and detail of the contours of the segmentations and choosing the most defined. The segmentation of each fish fillet image involves the detection of
    20 contours and regions of the image corresponding to anisakis.
    The luminance and saturation components contain all the information about the color. Depending on the value of the degree of color or hue, the domain of the color of each of the pixels is obtained. In the case that concerns us the luminance we can define it as the surface density of luminous flux that emerges from the fish fillet when it is illuminated. Saturation is defined as the chromaticity or intensity of color, so that for a constant degree of color, blue for example, different saturation values produce blue pixels, ranging from deep blue to weak blue. With a very low saturation, we have an achromatic pixel (black, white or gray), with an intensity of gray according to the
    30 luminance component.
    The method used to distinguish between chromatic and achromatic colors is based on a threshold of the saturation component. It is applied to segmentation by highlighting the chromatic pixels: for these pixels, the saturation is set to 255 and the luminance to 128. About a
    35 maximum value of 255:
    - If the saturation is greater than or equal to 20% and the luminance greater than or equal to 75%, the pixels are considered high brightness chromatic. -If the saturation is less than 20% and the luminance greater than or equal to 75% can be classified as white. -If the luminance is less than 25% the pixel is black.
    After carrying out the different segmentation and thresholding processes, an identification by form is carried out, for which an algorithm based on the correspondence of geometric patterns (Geometric Model Finder -GMF) is used compared to conventional pattern search technology by normalized gray-scale correlation, commonly called ~ NGCH.
    The algorithm developed recognizes the geometry of the anisakis using a series of boundary curves that do not correspond to a grid of pixels and then looks for similar shapes in the segmentation of the image without relying on specific gray levels. The result is an improvement in the ability to locate anisakis with high precision.
    The comparison with geometric forms (patterns) of anisakis is difficult and complex, since numerous variables can alter the way in which the possible anisakis appear in the image capture. The basic technology used in these cases is based on a pixel-grid analysis process (normalized correlation). This method to determine the XIY position looks for a similarity between a gray reference image of an anisakis and the objects in the captured image. This method limits both the ability to locate anisakis and the accuracy with which it can be located in conditions of variable appearance and changes in the angle, size and tones of the anisakis as they appear in fish fillets. To overcome these limitations, the team uses GMF technology that knows the geometry of an anisakis using a series of boundary curves that do not correspond to a pixel grid. After this process, look for similar shapes in the image without relying on specific levels of gray. The result is an improvement in the location of anisakis precisely despite changes in angle, size and tone.
    For the development and creation of the model based on the correspondence of geometric patterns (GMF) a pre-treatment of the images is carried out to eliminate noise and a thresholding is also carried out in order to obtain a binary image with the silhouette of the different configurations in which can be found anisakis in a fish fillet.
    Other processes carried out by the system are the classification of zones by model
    5 intensity detection in bands ranging from 230 to 700 nanometers. Hemodel processes and delivers intensity levels and number of pixels of each of thebands in which we section the mentioned frequency range.
    After obtaining the data from the previous algorithms a neural network,
    10 as shown in Figure 6, includes all the morphological, silhouette and color characteristics of each individual. The procedure of extraction of these characteristics allows the identification of an input vector to the neural network with more than 30 parameters that achieves a high efficiency after the training and validation of the network. Specifically, a set of more than 20 geometric parameters and 10 intensity parameters are used and
    15 presence in the determined wavelengths. Among the different parameters it is considered:
    o Total area / Total length.
    o Total area! Medium width
    o Total length! Medium width
    20 o Intensity per band.
    o Percentage per band.
    The modeling procedure follows an aggregative scheme. For this equipment, the base neuron function for the network has been identified. Then the
    25 structure (based on the multilayer perceptron architecture), conditioned by the number of input (characteristics) and output neurons.
    In order to offer an estimate of the level of prediction of the network, the segmentation of the numerical range of output of the network (between 0 and 1, or 0% and 100%) is made in three
    30 ranges, applying the previously established confidence index labeling: LOW (O-60%), MEDIUM (60 -80%) AND HIGH (80 -100%).
    In those cases in which more than two output neurons have similar parameters, the gradation is reduced by one level for the neuron with the highest value.
    35 15
    The training method used on the neural network is the Resillient Backpropagation (RPROP). This method is a variation within the family of backpropagation algorithms, based on the delta rule, and first-order optimization. The convergence and robustness of this algorithm is superior to other algorithms of the same family. RPROP uses independent parameters that control the speed with which the objective function is traversed for each of the neural network weights. RPROP is also not affected by the saturation of neurons in the neural network, since only the derivative is used to determine the direction in the weight update. Consequently, it converges faster than algorithms based only on
    10 backpropagation.
    权利要求:
    Claims (19)
    [1]
    1. Anisakis parasite detection system in fish fillets, where the fish fillet (1) to be inspected is deposited in an inspection area (2), characterized by 5 that the system comprises: -a lighting module (10) configured to subject the fish fillet (1) to a pulsed illumination; - image capture means (20) configured to automatically capture a color image (3) of the fish fillet (1) while being subjected to pulsed lighting 10; -a control module (30) configured to synchronize image capture (3)
    with pulsed lighting;-processing media (40) configured to:
    • apply a segmentation process on the captured image (3),
    15 obtaining a binary image (6) with the silhouette of different individuals that may correspond to anisakis (4) in the fish fillet (1);
    • make a geometric identification of each individual present in the binary image (6) by correspondence of geometric patterns, obtaining geometric parameters of each individual;
    20 • extract the characteristics of intensity and percentage of presence per spectral band of each individual, obtaining color parameters of each individual;
    • introduce an input vector with the geometric and coloration parameters obtained for each individual in an artificial neural network that establishes a relationship between the morphological, silhouette and color characteristics of each
    25 individual with the characteristics of anisakis, to estimate whether or not the individual is anisakis.
    [2]
    2. System according to claim 1, characterized in that the lighting module (10)
    It is configured to generate pulsed illumination in programmable periods with a 30 on duration between 0.1 to 1 second per pulse and a shutdown duration between
    [0]
     0.1 to 5 seconds per pulse.
    [3]
    3. System according to any of the preceding claims, characterized in that the lighting module (10) is configured to generate the pulsed illumination within a wavelength band between 230 and 700 nanometers.
    System according to any of the preceding claims, characterized in that thelighting module (10) is configured to generate pulsed lighting with acontrolled irradiance of up to 2 W / cm2.
    [5]
    5. System according to any of the preceding claims, characterized in that the
    10 control module (30) is configured to synchronize the image capture (3) with the illumination pressed according to the wavelength and irradiance programmed in the lighting module (10).
    [6]
    6. System according to any of the preceding claims, characterized in that for
    In order to obtain the coloring parameters, the processing means (40) are configured to classify the intensity in different color bands between 230 and 700 nanometers of wavelength.
    [7]
    7. System according to any of the preceding claims, characterized in that the
    20 processing means (40) are configured to obtain the positions of the anisakis identified in the fish and their size.
    [8]
    8. System according to any of the preceding claims, characterized in that the
    lighting module (10) is formed by a plurality of high power LED emitters.
    [9]
    9. System according to any of the preceding claims, characterized in that, for obtaining the binary image (6), the processing means (40) are configured to:
    30-extract, from the captured image (3) and by background segmentation techniques, the bottom (5) that belongs to the inspection area (2) not hidden by the fish fillet (1), obtaining a segmented image of the fish steak;
    - apply a thresholding process on said segmented image.
    [10]
    10. System according to any of the preceding claims, characterized in that the processing means (40) are configured to obtain the binary image (6)
    5 by any one of the following ways: -a segmentation by wavelength band; -a segmentation by luminance; -a saturation segmentation; -apply several of the previous segmentations and choice of segmentation that
    10 offers an image with the contours of the most defined individuals.
    [11]
    11. Anisakis parasite detection method in fish fillets, where the fish fillet (1) to be inspected is deposited in an inspection area (2), characterized in that the method comprises:
    15-send the fish fillet (1) to a pulsed illumination;
    - Automatically capture a color image (3) of the fish fillet (1) while it is subjected to pulsed illumination, the image capture (3) being synchronized with the pulsed illumination;
    - apply a segmentation process on the captured image (3), obtaining a binary image (6) with the silhouette of different individuals that may correspond to anisakis (4) in the fish fillet (1);
    - make a geometric identification of each individual present in the binary image (6) through a process of correspondence of geometric patterns, obtaining geometric parameters of each individual;
    25 - Extract the characteristics of intensity and percentage of presence per spectral band of each individual, obtaining color parameters of each individual;
    - introduce an input vector with the geometric and coloring parameters obtained for each individual in an artificial neural network that establishes a relationship between the morphological, silhouette and color characteristics of each individual with the
    30 characteristics of anisakis, to estimate whether or not the individual is anisakis.
    [12]
    12. Method according to claim 11, characterized in that the pulsed illumination is of programmable periods with an on duration between 0.1 to 1 second per pulse and an off duration between 0.1 to 5 seconds per pulse.
    [13]
    13. Method according to any of claims 11 to 12, characterized in that the
    5 pulsed illumination is within a wavelength band between 230 and 700nanometers
    [14]
    14. Method according to any of claims 11 to 13, characterized in that the pulsed illumination is with a controlled irradiance of up to 2 W / cm2.
    [15]
    fifteen. Method according to any of claims 11 to 14, characterized in that the synchronization of the capture of the image (3) with the pulsed illumination is carried out according to the wavelength and irradiance of the pulsed illumination.
    Method according to any of claims 11 to 15, characterized in that the intensity is classified in different color bands between 230 and 700 nanometers of wavelength for obtaining the coloration parameters.
    [17]
    17. Method according to any of claims 11 to 16, characterized in that
    20 comprises obtaining the positions of the anisakis identified in the fish and their size.
    [18]
    18. Method according to any of claims 11 to 17, characterized in that the pulsed lighting is performed using a plurality of high power LED emitters.
    [19]
    19. Method according to any of claims 11 to 18, characterized in that obtaining the binary image (6) comprises: - extracting, from the captured image (3) and by background segmentation techniques, the background (5) that belongs to the inspection area (2) not hidden by the fish fillet (1),
    30 obtaining a segmented image of the fish fillet; -apply a thresholding process on said segmented image.
    [20]
    20. Method according to any of claims 11 to 19, characterized in that the obtaining of the binary image (6) is carried out in any of the following ways:
    - a segmentation by wavelength band;
    5 -a segmentation by luminance;
    - a saturation segmentation;
    - apply several of the above segmentations and choose the segmentation offered
    an image with the contours of the most defined individuals.
    类似技术:
    公开号 | 公开日 | 专利标题
    Park et al.1998|Hyperspectral imaging for detecting fecal and ingesta contamination on poultry carcasses
    Anthonys et al.2009|An image recognition system for crop disease identification of paddy fields in Sri Lanka
    Zion et al.2007|Real-time underwater sorting of edible fish species
    Neal et al.2012|Measuring shape
    CN104880412B|2020-02-07|Freshness information output method and freshness information output device
    Wold et al.2001|Detection of parasites in cod fillets by using SIMCA classification in multispectral images in the visible and NIR region
    TWI645339B|2018-12-21|Non-invasive multimodal biometrical identification system of animals
    CN104034710B|2016-08-24|Plant disease detection method based on chlorophyll fluorescence and imaging technique
    JPWO2019151393A1|2020-12-03|Food inspection system, food inspection program, food inspection method and food production method
    US20200364862A1|2020-11-19|Wound imaging and analysis
    ES2552405B1|2016-09-14|System and method of detection of Anisakis parasites in fish fillets
    Yang et al.2019|Tea diseases detection based on fast infrared thermal image processing technology
    US10776667B2|2020-09-15|Speckle contrast analysis using machine learning for visualizing flow
    US10620417B2|2020-04-14|Method for generating a reflection-reduced contrast image and corresponding device
    CN109492714A|2019-03-19|Image processing apparatus and its method
    Waranusast et al.2016|Egg size classification on Android mobile devices using image processing and machine learning
    JP2018106720A|2018-07-05|Apparatus and method for image processing
    Coelho et al.2016|A machine vision system for automatic detection of parasites Edotea magellanica in shell-off cooked clam Mulinia edulis
    Park et al.2019|Line-scan imaging analysis for rapid viability evaluation of white-fertilized-egg embryos
    Abd Aziz et al.2020|A Review on Computer Vision Technology for Monitoring Poultry Farm—Application, Hardware, and Software
    Li2012|Hyperspectral imaging technology used in tongue diagnosis
    JP5860970B2|2016-02-16|Post-processing to improve eigenimage generation
    Lu et al.2015|Detecting defects on citrus surface based on circularity threshold segmentation
    Park2016|Quality Evaluation of Poultry Carcass
    Sankar et al.2013|White spot syndrome virus detection in shrimp images using image segmentation techniques
    同族专利:
    公开号 | 公开日
    ES2552405B1|2016-09-14|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    US6061086A|1997-09-11|2000-05-09|Canopular East Inc.|Apparatus and method for automated visual inspection of objects|
    US20070238147A1|2006-03-24|2007-10-11|Nippon Suisan Kaisha, Ltd.|Method of detecting foreign matter|
    US20120169863A1|2010-12-29|2012-07-05|Ido Bachelet|Apparatus and method for automatic detection of pathogens|WO2019121854A1|2017-12-20|2019-06-27|Intervet International B.V.|Method and system for external fish parasite monitoring in aquaculture|
    US10863727B2|2015-10-22|2020-12-15|Intervet Inc.|Method for automatic sea lice monitoring in salmon aquaculture|
    法律状态:
    2016-09-14| FG2A| Definitive protection|Ref document number: 2552405 Country of ref document: ES Kind code of ref document: B1 Effective date: 20160914 |
    优先权:
    申请号 | 申请日 | 专利标题
    ES201430789A|ES2552405B1|2014-05-27|2014-05-27|System and method of detection of Anisakis parasites in fish fillets|ES201430789A| ES2552405B1|2014-05-27|2014-05-27|System and method of detection of Anisakis parasites in fish fillets|
    [返回顶部]