CN113688959A - Plant disease and insect pest diagnosis method and system based on artificial intelligence - Google Patents
Plant disease and insect pest diagnosis method and system based on artificial intelligence Download PDFInfo
- Publication number
- CN113688959A CN113688959A CN202111249226.1A CN202111249226A CN113688959A CN 113688959 A CN113688959 A CN 113688959A CN 202111249226 A CN202111249226 A CN 202111249226A CN 113688959 A CN113688959 A CN 113688959A
- Authority
- CN
- China
- Prior art keywords
- plant
- plant sample
- labeled
- sample image
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 title claims abstract description 116
- 201000010099 disease Diseases 0.000 title claims abstract description 114
- 241000607479 Yersinia pestis Species 0.000 title claims abstract description 58
- 238000000034 method Methods 0.000 title claims abstract description 42
- 241000238631 Hexapoda Species 0.000 title claims abstract description 41
- 238000003745 diagnosis Methods 0.000 title claims abstract description 34
- 238000013473 artificial intelligence Methods 0.000 title claims abstract description 23
- 238000012549 training Methods 0.000 claims abstract description 23
- 241000196324 Embryophyta Species 0.000 claims description 413
- 230000006870 function Effects 0.000 claims description 47
- 230000004927 fusion Effects 0.000 claims description 21
- 239000011159 matrix material Substances 0.000 claims description 16
- 238000010586 diagram Methods 0.000 claims description 14
- 238000005286 illumination Methods 0.000 claims description 11
- 238000003708 edge detection Methods 0.000 claims description 8
- 208000024891 symptom Diseases 0.000 claims description 7
- 238000012545 processing Methods 0.000 claims description 6
- 230000011218 segmentation Effects 0.000 claims description 5
- 230000005012 migration Effects 0.000 claims description 4
- 238000013508 migration Methods 0.000 claims description 4
- 238000004590 computer program Methods 0.000 claims description 3
- 208000015181 infectious disease Diseases 0.000 abstract description 2
- 230000008569 process Effects 0.000 description 6
- 238000013528 artificial neural network Methods 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 239000000243 solution Substances 0.000 description 2
- 241000238876 Acari Species 0.000 description 1
- 208000035143 Bacterial infection Diseases 0.000 description 1
- 241000237858 Gastropoda Species 0.000 description 1
- 208000031888 Mycoses Diseases 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 230000012010 growth Effects 0.000 description 1
- 238000002347 injection Methods 0.000 description 1
- 239000007924 injection Substances 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 238000007500 overflow downdraw method Methods 0.000 description 1
- 230000008635 plant growth Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Biophysics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Probability & Statistics with Applications (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Evolutionary Biology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Investigating Or Analysing Biological Materials (AREA)
Abstract
The invention relates to a plant disease and insect pest diagnosis method and system based on artificial intelligence, and belongs to the technical field of plant disease and insect pest diagnosis. The method comprises the following steps: obtaining disease style characteristics of each plant sample image with the label according to the color texture distribution map; training a variational self-encoder by utilizing each plant sample image with a label to obtain a Gaussian model corresponding to each plant sample image to be labeled; according to the Gaussian model, distributing labels for the plant sample images to be labeled to obtain label data corresponding to the plant sample images to be labeled; training a target network according to the plant sample images with labels, the label data corresponding to the plant sample images with labels, the plant sample images to be labeled and the label data corresponding to the plant sample images to be labeled; and inputting the target plant image into the trained target network to obtain a diagnosis result corresponding to the target plant image. The invention can improve the accuracy of diagnosing plant infection diseases and insect pests by the target network.
Description
Technical Field
The invention relates to the technical field of plant disease and insect pest diagnosis, in particular to a plant disease and insect pest diagnosis method and system based on artificial intelligence.
Background
Plant diseases and insect pests refer to diseases and insect pests infected during the growth of plants, the diseases and insect pests of the plants are mainly fungal diseases, bacterial diseases and the like, the insect pests of the plants are mainly insects, mites, snails and the like, and the infected plant diseases and insect pests of the plants seriously affect the normal growth and results of the plants and may also cause the death of the plants.
The existing plant disease and insect pest diagnosis method is generally used for diagnosing whether plant diseases and insect pests occur or not based on manual or neural networks, the mode based on manual diagnosis is complicated and omission may occur, and the mode based on neural network diagnosis needs a professional specially knowing plant diseases and insect pests to label, because the labeling cost and energy are considered, the scale of the finally obtained data set with labels is not large, and the accuracy of the mode for diagnosing plant diseases and insect pests based on neural network is low.
Disclosure of Invention
The invention provides a plant disease and insect pest diagnosis method and system based on artificial intelligence, which are used for solving the problem that plant disease and insect pest cannot be accurately diagnosed in the prior art, and adopt the following technical scheme:
in a first aspect, an embodiment of the present invention provides an artificial intelligence-based plant disease and pest diagnosis method and system, comprising the following steps:
acquiring a plant sample image set with a label, a plant sample image set to be labeled and a target plant image; the labeled plant sample image set comprises plant sample images of different state categories with labels; the different status categories comprise different pest categories and pest-free categories;
obtaining a color texture distribution diagram corresponding to each plant sample image with the label according to each plant sample image with the label; obtaining disease style characteristics of each plant sample image with the label according to the color texture distribution map;
training a variational self-encoder by utilizing the plant sample images with the labels to obtain Gaussian models corresponding to the plant sample images to be labeled; the loss function of the variation self-encoder is constructed by the disease style characteristics of each plant sample image with labels and a Gaussian model; the variation self-encoder is used for extracting disease style characteristics of each plant sample image with a label and distributing the label for the plant sample image to be labeled;
according to the Gaussian model, distributing labels for the plant sample images to be labeled to obtain label data corresponding to the plant sample images to be labeled;
training a target network according to each plant sample image with a label, label data corresponding to each plant sample image with the label, each plant sample image to be labeled and the label data corresponding to each plant sample image to be labeled, wherein the target network is used for diagnosing whether the plant is infected with the plant diseases and insect pests and the corresponding specific plant disease and insect pest type when the plant is infected with the plant diseases and insect pests;
and inputting the target plant image into the trained target network to obtain a diagnosis result corresponding to the target plant image.
The invention also provides a plant disease and insect pest diagnosis system based on artificial intelligence, which comprises a memory and a processor, wherein the processor executes a computer program stored in the memory so as to realize the plant disease and insect pest diagnosis method based on artificial intelligence.
According to the color texture distribution map, acquiring disease style characteristics of each plant sample image with labels; training a variational self-encoder by utilizing each plant sample image with a label to obtain a Gaussian model corresponding to each plant sample image to be labeled; according to the Gaussian model, distributing labels for the plant sample images to be labeled to obtain label data corresponding to the plant sample images to be labeled; training a target network according to the plant sample images with labels, the label data corresponding to the plant sample images with labels, the plant sample images to be labeled and the label data corresponding to the plant sample images to be labeled; and inputting the target plant image into the trained target network to obtain a diagnosis result corresponding to the target plant image. According to the invention, each plant sample image with a label is used as a basis for training the variational self-encoder, the trained variational self-encoder is used as a basis for obtaining label data corresponding to each plant sample image to be labeled, and each plant sample image with a label, the label data corresponding to each plant sample image with a label, each plant sample image to be labeled and the label data corresponding to each plant sample image to be labeled are used as a basis for training the target network.
Preferably, the method for obtaining the color texture distribution map corresponding to each labeled plant sample image according to each labeled plant sample image comprises:
removing the illumination information and the texture information on each plant sample image with the label to obtain an RGB image without illumination and texture corresponding to each plant sample image with the label;
segmenting the RGB image without illumination and texture by utilizing a superpixel segmentation algorithm to obtain a color distribution map corresponding to each labeled plant sample image;
carrying out edge detection on the plant sample images with the labels by using an edge detection algorithm to obtain edge distribution maps corresponding to the plant sample images with the labels;
and obtaining a color texture distribution diagram corresponding to each labeled plant sample image according to the color distribution diagram and the edge gray level distribution diagram.
Preferably, the method for obtaining the disease style characteristics of each labeled plant sample image according to the color texture distribution map comprises the following steps:
processing the color texture distribution map by using a style migration algorithm to obtain a Gram matrix corresponding to each color texture distribution map;
and taking the Gram matrix as a disease style characteristic corresponding to each plant sample image with the label.
Preferably, the method for training the variational self-encoder by using the labeled plant sample images comprises the following steps:
and respectively inputting the plant sample images with the labels into a variation self-encoder, and training the variation self-encoder by utilizing a reconstruction loss function, an intra-class disease style loss function, an inter-class disease style loss function and a global loss function.
Preferably, the reconstruction loss function is:
wherein,is as followsTagged second in individual status categoryThe variation corresponding to each plant sample image is from the reconstruction loss function of the encoder,is corresponding to plantTagged second in individual status categoryA Gaussian model corresponding to the image of each plant sample,is corresponding to plantThe gaussian model corresponding to each state class,is corresponding to plantTagged second in individual status categoryAn image of a sample of an individual plant,corresponding to plants output from the decoder of the encoder for variationTagged second in individual status categoryA reconstructed image corresponding to the image of the plant sample,is composed ofAndis/are as followsDivergence.
Preferably, the intra-category-disease style loss function is:
wherein,is as followsTagged second in individual status categoryThe variation corresponding to each plant sample image is from the intra-category disease style loss function of the encoder,is corresponding to plantThe number of labeled plant specimen images in each status category,is corresponding to plantTagged second in individual status categoryAn image of a sample of an individual plant,is corresponding to plantTagged second in individual status categoryAn image of a sample of an individual plant,is corresponding to plantTagged second in individual status categoryThe plant sample image and the labeled secondThe degree of fusion of disease style characteristics of individual plant sample images,is as followsTagged second in individual status categoryThe plant sample image and the labeled secondThe degree of gaussian model fusion between individual plant sample images,。
preferably, the inter-class disorder style loss function is:
wherein,is as followsTagged second in individual status categoryThe variation corresponding to each plant sample image is from the inter-class symptom style loss function of the encoder,the number of state categories corresponding to the plants,is corresponding to plantThe status of each of the plurality of status categories,is corresponding to plantThe status of each of the plurality of status categories,,is corresponding to plantStatus class andthe difficulty of gaussian model discrimination between individual state classes,is corresponding to plantStatus class andthe difficulty of distinguishing the disease style characteristics between individual status categories.
Preferably, the global penalty function is:
wherein,is corresponding to plantTagged second in individual status categoryThe variation corresponding to each plant sample image is from the global loss function of the encoder,is corresponding to plantThe gaussian model corresponding to each state class,is a standard Gaussian model and is used as a model,is composed ofAndis/are as followsDivergence.
Preferably, the method for obtaining the label data corresponding to the plant sample image to be labeled comprises:
inputting each plant sample image to be labeled into a trained variational self-encoder to obtain a Gaussian model corresponding to each plant sample image to be labeled;
calculating the probability that each plant sample image to be labeled belongs to each state category corresponding to the plant according to the Gaussian model of each plant sample image to be labeled;
and taking the probability that each plant sample image to be labeled belongs to each state type corresponding to the plant as label data corresponding to the plant sample image to be labeled.
Drawings
To more clearly illustrate the embodiments of the present invention or the technical solutions and advantages of the prior art, the following description will be made
While the drawings necessary for describing the embodiments or prior art are briefly described, it should be apparent that the drawings in the following description are merely examples of the invention and that other drawings may be derived from those drawings by those of ordinary skill in the art without inventive step.
FIG. 1 is a flow chart of the plant disease and insect pest diagnosis method based on artificial intelligence.
Detailed Description
The technical solutions in the embodiments of the present invention will be described clearly and completely with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, rather than all embodiments, and all other embodiments obtained by those skilled in the art based on the embodiments of the present invention belong to the protection scope of the embodiments of the present invention.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
The embodiment provides a plant disease and insect pest diagnosis method based on artificial intelligence, which is described in detail as follows:
as shown in figure 1, the plant disease and insect pest diagnosis method based on artificial intelligence comprises the following steps:
s001, acquiring a labeled plant sample image set, a plant sample image set to be labeled and a target plant image; the labeled plant sample image set comprises plant sample images of different state categories with labels; the different status categories include different pest categories and a pest-free category.
In this embodiment, the unmanned aerial vehicle carries the RGB high-definition camera to acquire images of plants in different growing periods in different areas, the plant images acquired by the unmanned aerial vehicle carrying the RGB high-definition camera in a historical time period are recorded as a plant sample image set, and the plant images acquired by the unmanned aerial vehicle carrying the RGB high-definition camera in real time are recorded as target plant images.
In the embodiment, when the unmanned aerial vehicle carries an RGB high-definition camera to acquire images, the unmanned aerial vehicle flies in low altitude; in this embodiment, the length of the historical time period and the frame rate of image acquisition are set according to actual conditions, but the acquired plant sample image set is large-scale and the plant sample image set includes all plant diseases and insect pests.
In this embodiment, a part of plant sample images in the collected plant sample image set is labeled manually, so that a labeled plant sample image set is obtained, and the remaining part of plant sample images in the collected plant sample image set is marked as a plant sample image set to be labeled; the labeled plant sample image set comprises plant sample images of different state categories with labels; the different status categories comprise different pest categories and pest-free categories; in this embodiment, the same label data is labeled on the plant sample images infected with the same pest and disease category, and the same label data is labeled on all the plant sample images which are normal, i.e., not infected with pest and disease, for example, the plant sample images infected with yellow leaf disease are all labeled as 1; the types of pests and diseases contained in the labeled plant sample image set are complete.
S002, obtaining a color texture distribution map corresponding to each plant sample image with the label according to each plant sample image with the label; and obtaining the disease style characteristics of each plant sample image with the label according to the color texture distribution map.
In the embodiment, the illumination information on each plant sample image with the label is removed by processing each plant sample image with the label, so that the influence of illumination on subsequent analysis is avoided; the specific method comprises the following steps: inputting each plant sample image with a label into a RetinexNet network to obtain a non-illuminated RGB image corresponding to each plant sample image with the label, wherein the RetinexNet network is a known prior art, and therefore, the embodiment is not described in detail.
In this embodiment, a 5 × 5 gaussian kernel is used to perform blurring processing on the non-illuminated RGB images corresponding to the labeled plant sample images, and texture information on the non-illuminated RGB images corresponding to the labeled plant sample images is removed to obtain non-illuminated non-textured RGB images corresponding to the labeled plant sample images; then, segmenting the RGB image without illumination and texture corresponding to each plant sample image with the label according to an SLIC superpixel segmentation algorithm to obtain a plurality of regions, wherein the color in each region is similar; in this embodiment, the pixel value of each pixel point in the same region is set as the average value of the summed pixel values in the region, so that the RGB image without illumination and texture corresponding to each labeled plant sample image containing color patches of different colors can be obtained; and marking the RGB image without illumination and texture corresponding to each labeled plant sample image containing different color patches as a color distribution diagram corresponding to each labeled plant sample image, wherein the color distribution diagram can indicate the color distribution condition of the plant after the plant is infected with the plant diseases and insect pests.
In this embodiment, the fuzzy processing and SLIC superpixel segmentation algorithms are well-known technologies, and therefore this embodiment is not described in detail; as another embodiment, other ways of blurring the non-illuminated RGB images corresponding to the labeled plant sample images or segmenting the non-illuminated non-textured RGB images corresponding to the labeled plant sample images by using other segmentation algorithms may be selected according to actual conditions.
In this embodiment, graying is performed on each labeled plant sample image to obtain a gray image of each labeled plant sample, then a Canny edge detection algorithm is used to perform edge extraction on the obtained gray image of each labeled plant sample to obtain an edge distribution map corresponding to each labeled plant sample gray image, and the edge distribution map can indicate texture distribution conditions after plant infection by plant diseases and insect pests; in this embodiment, the Canny edge detection algorithm is a known technique, and therefore this embodiment is not described in detail; as other embodiments, other algorithms can be used to perform edge extraction on the surface image of the injection molded part according to different requirements, such as a Sobel edge detection algorithm or a Roberts edge detection algorithm.
In this embodiment, the color distribution map corresponding to each plant sample image with a label and the edge distribution map corresponding to each plant sample gray-scale image with a label are fused to obtain the color texture distribution map corresponding to each plant sample image with a label; the specific fusion method comprises the following steps: copying three marks from the edge distribution diagram corresponding to the gray scale image of each plant sample with the single channel and carrying out weighting summation on the edge distribution diagram corresponding to the three channels of each plant sample image with the label and the color distribution diagram corresponding to each plant sample image with the label, and marking the image after weighting summation as the color texture distribution diagram corresponding to each plant sample image with the label.
In this embodiment, the weight of the three channels of the edge distribution map corresponding to each labeled plant sample image in the weighted summation process is 0.4, and the weight of the color distribution map corresponding to each labeled plant sample image is 0.6; as another embodiment, different weights may be assigned to the edge distribution map of the three channels corresponding to the labeled plant sample images and the color distribution map corresponding to the labeled plant sample images according to the actual situation, but the sum of the weights may be 1.
In this embodiment, the obtained color texture distribution map corresponding to each plant sample image with a label is analyzed to obtain a disease style characteristic corresponding to each plant sample image with a label, and the obtained disease style characteristic is used as a condition for constructing a loss function in the subsequent training of a variational self-encoder; the specific analysis method comprises the following steps: processing the obtained color texture distribution map corresponding to each plant sample image with the label by using a style migration algorithm to obtain a Gram matrix of the color texture distribution map corresponding to each plant sample image with the label, and taking the Gram matrix of the color texture distribution map corresponding to each plant sample image with the label as a disease style characteristic corresponding to each plant sample image with the label; the style migration algorithm is a well-known technique and thus the embodiment will not be described in detail. In this embodiment, the color texture distribution map obtained by the above process eliminates the mottle on each plant sample image with a label and enhances the disease style characteristics of each plant sample image with a label, as compared with each plant sample image with a label.
Step S003, training a variational self-encoder by utilizing the plant sample images with the labels to obtain Gaussian models corresponding to the plant sample images to be labeled; the loss function of the variation self-encoder is constructed by the disease style characteristics of each plant sample image with labels and a Gaussian model; the variation self-encoder is used for extracting disease style characteristics of each labeled plant sample image and distributing a label for the plant sample image to be labeled.
In the embodiment, the trained variational self-encoder is used as a basis for subsequently distributing labels for each plant sample image to be labeled; in this embodiment, a variational self-encoder is first constructed, and the structure of the variational self-encoder is a known technology, so this embodiment is not described in detail; the working principle of the variational self-encoder is as follows: inputting an image into a variational self-encoder, outputting a corresponding output from the encoder of the variational self-encoder for the inputted imageThe matrix is a matrix of a plurality of matrices,the matrix is represented asHigh dimensional of dimensional spaceAnd is andfirst row representation of the matrixMean of the dimensional gaussian model, the second row representing the elements on the diagonal of the gaussian covariance matrix; sampling one from the corresponding Gaussian model of the imageOne of the spaceDimension vector of theThe dimension vector input variation obtains a reconstructed image corresponding to the image from a decoder of an encoder.
In this embodiment, in the variational self-encoder128, as another embodiment, may be used according to the actual situationA different value is set, which may be 256, for example.
In this embodiment, each labeled plant sample image corresponding to each state category is input into a variation self-encoder, and the variation self-encoder is supervised and iteratively trained by using a conventional loss function, an intra-class disease style loss function, an inter-class disease style loss function, and a global loss function, for example, for the second plant corresponding to the plantTagged second in individual status categoryThe conventional loss function for individual plant sample images is:
wherein,is as followsTagged second in individual status categoryThe variation corresponding to each plant sample image is from the reconstruction loss function of the encoder,is corresponding to plantTagged second in individual status categoryA Gaussian model corresponding to the image of each plant sample,is corresponding to plantThe gaussian model corresponding to each state class,is corresponding to plantTagged second in individual status categoryAn image of a sample of an individual plant,corresponding to plants output from the decoder of the encoder for variationTagged second in individual status categoryA reconstructed image corresponding to the image of the plant sample,is composed ofAndis/are as followsDivergence;is composed ofThe L2 norm; this embodiment requires that the input of the variational autocoder be close to the output, i.e.Approach toThen, thenThe requirement approaches 0.
In the present embodiment, the first and second electrodes are,by corresponding of the plantGaussian model corresponding to each state classThe mean and covariance matrix of (a), analyzing the plant's corresponding secondGaussian model corresponding to each state class(ii) a Corresponding to plantsGaussian model corresponding to each state classThe mean and covariance matrices of (a) are:
wherein,is corresponding to plantGaussian model corresponding to each state classThe average value of (a) of (b),is corresponding to plantThe number of labeled plant specimen images in each status category,is as followsTagged second in individual status categoryGaussian model corresponding to plant sample imageThe mean value of (a);is corresponding to plantThe covariance matrix corresponding to each state class,is as followsTagged second in individual status categoryGaussian model corresponding to plant sample imageThe covariance matrix of (2).
In this embodiment, the process can be used to obtain the second plantA state class pairCorresponding Gaussian modelIs the mean value ofMean value of Gaussian model means corresponding to all labeled plant sample images in each state category, and mean value of Gaussian model means corresponding to plantThe covariance matrix corresponding to each state class isMean of covariance matrices corresponding to all labeled plant sample images in each state class, henceGaussian model corresponding to each state classIs as followsAnd (3) the distribution characteristics of all labeled plant sample images in the state category in a high-dimensional space.
In this example, the plants correspond toTagged second in individual status categoryThe intra-category-style loss function for individual plant sample images is:
wherein,is as followsTagged second in individual status categoryThe variation corresponding to each plant sample image is from the intra-category disease style loss function of the encoder,is corresponding to plantThe number of labeled plant specimen images in each status category,is corresponding to plantTagged second in individual status categoryAn image of a sample of an individual plant,is corresponding to plantTagged second in individual status categoryAn image of a sample of an individual plant,is corresponding to plantTagged second in individual status categoryThe plant sample image and the labeled secondThe degree of fusion of disease style characteristics of individual plant sample images,is as followsTagged second in individual status categoryThe plant sample image and the labeled secondThe degree of gaussian model fusion between individual plant sample images,。
in this embodiment, the followingImage of plant sample andsimilarity of disease style characteristics between individual plant sample images is obtainedTagged second in individual status categoryThe plant sample image and the labeled secondThe degree of fusion of disease pattern characteristics of individual plant sample images, andimage of plant sample andsimilarity of disease style characteristics between individual plant sample images and the firstTagged second in individual status categoryThe plant sample image and the labeled secondThe disease style characteristic fusion degrees of the plant sample images form a positive correlation; and calculating the corresponding second of the plants according to the following formulaTagged second in individual status categoryThe plant sample image and the labeled secondDisease style characteristic fusion degree of individual plant sample image:
Wherein,is corresponding to plantTagged second in individual status categoryThe plant sample image and the labeled secondThe degree of fusion of disease style characteristics between individual plant sample images,is as followsTagged second in individual status categoryThe disease style characteristics corresponding to the images of the individual plant samples,is as followsTagged second in individual status categoryDisease style characteristics corresponding to the plant sample images;is as followsImage of plant sample andthe difference in the disease style characteristics between individual plant sample images,is composed ofThe L2 norm of (a),is as followsImage of plant sample andsimilarity of disease style characteristics between individual plant sample images; when in useThe smaller the value of (a) is,the smaller the value of the L2 norm, the correspondingThe greater the value of (A), i.e. the firstImage of plant sample andthe higher the similarity of disease style characteristics among the plant sample images;is at the firstIn the individual state classAn image of a plant sample andthe sum of the similarity of the disease style characteristics of all the labeled plant sample images,is a normalization coefficient.
In the present embodiment, the first and second electrodes are,the larger the value of (A) is, theThe first in the individual state classImage of plant sample andthe greater the degree of fusion of disease pattern characteristics of the individual plant sample images, and the requirements of this embodiment are on the second placeIn the individual state classGaussian model of plant sample image andthe Gaussian models of the plant sample images are relatively close to each other, namely the corresponding Gaussian model fusion degree is larger, so that for the same state class, the variational self-encoder can classify the two disease symptoms with large style feature similarity into one state class;smaller value of (A) indicates thatThe first in the individual state classImage of plant sample andthe smaller the degree of fusion of disease pattern characteristics of the individual plant sample images, and the embodiment requiresGaussian model of plant sample image andthe Gaussian models of the plant sample images are relatively far away from each other, namely the corresponding Gaussian models are less in fusion degree, so that for the same state class, the variational self-encoder can classify the two disease style characteristics with small similarity into the same state class.
In the present embodiment, the method is as followsIn the individual state classImage of plant sample andthe similarity of Gaussian models between the images of the individual plant samples is obtainedTagged second in individual status categoryThe plant sample image and the labeled secondThe degree of Gaussian model fusion between individual plant sample images; according to the following formulaTagged second in individual status categoryThe plant sample image and the labeled secondThe degree of gaussian model fusion between individual plant sample images:
wherein,is as followsTagged second in individual status categoryThe plant sample image and the labeled secondThe degree of gaussian model fusion between individual plant sample images,is as followsTagged second in individual status categoryThe mean value of the gaussian model corresponding to each plant sample image,is as followsTagged second in individual status categoryThe mean value of the gaussian model corresponding to each plant sample image,is composed ofThe L2 norm of (a),is composed ofAndthe similarity of the gaussian models of (1),the smaller the norm of L2 of (a),andthe greater the similarity of the gaussian models of (c),is at the firstIn the individual state classThe sum of the similarity of the Gaussian models of the plant sample images and all the labeled plant sample images,is a normalization coefficient.
In the present embodiment, the first and second electrodes are,greater values of (A) indicate inTagged second in individual status categoryThe plant sample image and the labeled secondThe more the Gaussian model fusion degree among the plant sample images is, the closer the Gaussian models are, and the distribution consistency of the Gaussian models is large; if it isSmaller value of (A) indicates thatTagged second in individual status categoryThe plant sample image and the labeled secondThe smaller the Gaussian model fusion degree among the plant sample images is, the closer the Gaussian models are, and the distribution consistency of the Gaussian models is small; required in this exampleApproaching 0, i.e.Approaching 0, i.e.The value of (a) is close to 0, so that the distribution consistency of the Gaussian models of the labeled plant sample images with large disease style feature fusion degree is large, the distribution consistency of the Gaussian models of the labeled plant sample images with small disease style feature fusion degree is small, and the variational self-encoder can have stronger feature extraction and characterization capabilities.
In this example, the plants correspond toTagged second in individual status categoryThe inter-class symptom style loss function for each plant sample image is:
wherein,is as followsTagged second in individual status categoryThe variation corresponding to each plant sample image is from the inter-class symptom style loss function of the encoder,the number of state categories corresponding to the plants,is corresponding to plantThe status of each of the plurality of status categories,is corresponding to plantThe status of each of the plurality of status categories,,is corresponding to plantStatus class andthe difficulty of gaussian model discrimination between individual state classes,is corresponding to plantStatus class andthe difficulty of distinguishing the disease style characteristics between individual status categories.
In this embodiment, the first step is calculatedStatus class anddifficulty in differentiating disease style characteristics between individual status categoriesThe specific method comprises the following steps: first, calculate theImage of each plant sample with label in individual state category and its secondSimilarity of disease style characteristics between the plant sample images with labels in the individual state category is obtainedStatus class andthe similarity set of the disease style characteristics among the individual state categories takes the maximum value in the similarity set of the disease style characteristics as the firstStatus class anddifficulty in differentiating disease style characteristics between individual status categories(ii) a When in useThe larger the value of (A) indicates that the similarity of the disease style characteristics corresponding to the two labeled plant sample images corresponding to the maximum value in the disease style characteristic similarity set is higher, the higher the similarity of the disease style characteristics corresponding to the two labeled plant sample images is, theStatus class andthe greater the difficulty of distinguishing between the individual status categories, and when the disease style characteristics of the two labeled plant specimen images are too similar, this can result in a difficulty for the variational self-encoder to distinguish the two labeled plant specimen images into two pest types.
In this embodiment, in order to enable the variational self-encoder to distinguish the labeled plant sample images with too similar disease symptoms style characteristics between different state types, it is required that the gaussian models corresponding to the labeled plant sample images with too similar disease symptoms style characteristics between different state types have a relatively large differentiability. Namely, the difficulty in distinguishing the Gaussian models corresponding to the labeled plant sample images with too similar disease style characteristics among different state types is small, so that the variational self-encoder can extract different characteristics of the two labels, and the distinguishing capability of the variational self-encoder is improved.
In this embodiment, the followingStatus class andthe difference degree of the Gaussian models among the individual state categories is obtained to obtain the corresponding second state of the plantStatus class anddifficulty of gaussian model discrimination between individual state classes; according to the following formulaStatus class andthe difficulty of gaussian model discrimination between individual state classes:
wherein,is as followsStatus class andthe difficulty of gaussian model discrimination between individual state classes,is as followsThe mean of the gaussian models corresponding to each state class,is as followsThe mean of the gaussian models corresponding to each state class,is composed ofThe L2 norm of (a),is as followsStatus class andthe degree of gaussian model difference between individual state classes,the larger the description isStatus class andthe less difficult the Gaussian model corresponding to each state class is to distinguish, andis a normalized coefficient.
In this example, the requirementsApproaching to 0, and making the Gaussian model corresponding to the labeled plant sample image with too similar disease style characteristics among different state types have smaller distinguishing difficulty.
In this example, the plants correspond toTagged second in individual status categoryThe global loss function for each plant sample image is:
wherein,is as followsTagged second in individual status categoryThe variation corresponding to each plant sample image is from the global loss function of the encoder,is corresponding to plantThe gaussian model corresponding to each state class,is a standard Gaussian model and is used as a model,is composed ofAndis/are as followsDivergence.
In the present embodiment, the first and second electrodes are,is a standard Gaussian model with the mean value of 0 vector and all diagonal elements of the covariance matrix of 1,is composed ofAndis/are as followsDivergence, this example requires the corresponding second of the plantAnd the Gaussian models corresponding to the state classes are distributed on the standard Gaussian model.
In this embodiment, the followingTagged second in individual status categoryThe reconstruction loss function of the variational self-encoder corresponding to each plant sample imageTagged second in individual status categoryVariation self-encoder intra-disease-like style loss function corresponding to individual plant sample imageTagged second in individual status categoryInter-class disease style loss function of variation self-encoder corresponding to plant sample image and second class disease style loss function corresponding to plantTagged second in individual status categoryObtaining the global loss function of the variational self-encoder corresponding to the plant sample image to obtain the second loss function corresponding to the plantTagged second in individual status categoryLoss values corresponding to individual plant sample images; calculating the corresponding second of the plants according to the following formulaTagged second in individual status categoryLoss values corresponding to individual plant sample images:
wherein,is corresponding to plantTagged second in individual status categoryLoss values corresponding to individual plant sample images.
In this embodiment, each plant sample image with a label according to the above process may obtain a corresponding loss value, and then the parameters of the variational self-encoder are updated by using a random gradient descent algorithm until convergence, thereby completing the training process of the variational self-encoder.
And step S004, distributing labels to the plant sample images to be labeled according to the Gaussian model to obtain label data corresponding to the plant sample images to be labeled.
In this embodiment, a plant sample image set to be labeled is input into a trained variational self-encoder, a gaussian model corresponding to each plant sample image to be labeled in the plant sample image set to be labeled is obtained, then, according to the corresponding gaussian model similarity between each plant sample image to be labeled and each state category corresponding to the plant sample image, the probability that each plant sample image to be labeled belongs to each state category corresponding to the plant is obtained, and the probability that each plant sample image to be labeled belongs to each state category corresponding to the plant is calculated according to the following formula:
wherein,is as followsThe plant sample image to be labeled belongs to the second image corresponding to the plantThe probability of the individual state class,is as followsA Gaussian model corresponding to the plant sample graph to be labeled,is as followsThe gaussian model corresponding to each state class,is composed ofAndis/are as followsThe divergence of the light beam is measured by the light source,the smallerAndthe higher the corresponding gaussian model similarity,is composed ofAndcorresponding gaussian model similarity; when in useThe larger the value of (A), the corresponding secondThe plant sample image to be labeled belongs to the second image corresponding to the plantThe greater the probability of a state class, whereIs a normalized coefficient.
In this embodiment, the probability that each plant sample image to be labeled belongs to each state category corresponding to the plant is used as the label data corresponding to the plant sample image to be labeled; therefore, through the process, the label can be distributed to each plant sample image to be labeled, and label data corresponding to each plant sample image to be labeled is obtained.
Step S005, training a target network according to the plant sample images with the labels, the label data corresponding to the plant sample images with the labels, the plant sample images to be labeled and the label data corresponding to the plant sample images to be labeled, wherein the target network is used for diagnosing whether the plants are infected with the plant diseases and insect pests and the corresponding specific plant disease types when the plants are infected with the plant diseases and insect pests.
In the embodiment, a target network is constructed, the target network is a fully-connected network, the labeled plant sample images, the label data corresponding to the labeled plant sample images, the plant sample images to be labeled and the label data corresponding to the plant sample images to be labeled are input into the target network for training, and the target network is trained by using a cross entropy loss function and a random gradient descent algorithm; the target network is used for diagnosing whether the plant is infected by the plant diseases and insect pests and the corresponding specific plant disease and insect pest types when the plant is infected by the plant diseases and insect pests.
And S006, inputting the target plant image into the trained target network to obtain a diagnosis result corresponding to the target plant image.
In this embodiment, the target plant image is input to the trained target network, the output result is the probability corresponding to each label, and the state class corresponding to the maximum probability value that the plant sample image to be labeled belongs to each state class corresponding to the plant is used as the state class of the plant sample image to be labeled.
According to the color texture distribution map, acquiring disease style characteristics of each plant sample image with labels; training a variational self-encoder by utilizing each plant sample image with a label to obtain a Gaussian model corresponding to each plant sample image to be labeled; according to the Gaussian model, distributing labels for the plant sample images to be labeled to obtain label data corresponding to the plant sample images to be labeled; training a target network according to the plant sample images with labels, the label data corresponding to the plant sample images with labels, the plant sample images to be labeled and the label data corresponding to the plant sample images to be labeled; and inputting the target plant image into the trained target network to obtain a diagnosis result corresponding to the target plant image. In this embodiment, each plant sample image with a label is used as a basis for training the variational self-encoder, the trained variational self-encoder is used as a basis for obtaining label data corresponding to each plant sample image to be labeled, and each plant sample image with a label, the label data corresponding to each plant sample image with a label, each plant sample image to be labeled and the label data corresponding to each plant sample image to be labeled are used as a basis for training the target network.
The plant disease and insect pest diagnosis system based on artificial intelligence comprises a memory and a processor, wherein the processor executes a computer program stored in the memory to realize the plant disease and insect pest diagnosis method based on artificial intelligence.
It should be noted that the order of the above-mentioned embodiments of the present invention is merely for description and does not represent the merits of the embodiments, and in some cases, actions or steps recited in the claims may be executed in an order different from the order of the embodiments and still achieve desirable results.
Claims (10)
1. A plant disease and insect pest diagnosis method based on artificial intelligence is characterized by comprising the following steps:
acquiring a plant sample image set with a label, a plant sample image set to be labeled and a target plant image; the labeled plant sample image set comprises plant sample images of different state categories with labels; the different status categories comprise different pest categories and pest-free categories;
obtaining a color texture distribution diagram corresponding to each plant sample image with the label according to each plant sample image with the label; obtaining disease style characteristics of each plant sample image with the label according to the color texture distribution map;
training a variational self-encoder by utilizing the plant sample images with the labels to obtain Gaussian models corresponding to the plant sample images to be labeled; the loss function of the variation self-encoder is constructed by the disease style characteristics of each plant sample image with labels and a Gaussian model; the variation self-encoder is used for extracting disease style characteristics of each plant sample image with a label and distributing the label for the plant sample image to be labeled;
according to the Gaussian model, distributing labels for the plant sample images to be labeled to obtain label data corresponding to the plant sample images to be labeled;
training a target network according to each plant sample image with a label, label data corresponding to each plant sample image with the label, each plant sample image to be labeled and the label data corresponding to each plant sample image to be labeled, wherein the target network is used for diagnosing whether the plant is infected with the plant diseases and insect pests and the corresponding specific plant disease and insect pest type when the plant is infected with the plant diseases and insect pests;
and inputting the target plant image into the trained target network to obtain a diagnosis result corresponding to the target plant image.
2. A method according to claim 1, wherein the method for obtaining a color texture distribution map corresponding to each labeled plant sample image from each labeled plant sample image comprises:
removing the illumination information and the texture information on each plant sample image with the label to obtain an RGB image without illumination and texture corresponding to each plant sample image with the label;
segmenting the RGB image without illumination and texture by utilizing a superpixel segmentation algorithm to obtain a color distribution map corresponding to each labeled plant sample image;
carrying out edge detection on the plant sample images with the labels by using an edge detection algorithm to obtain edge distribution maps corresponding to the plant sample images with the labels;
and obtaining a color texture distribution diagram corresponding to each labeled plant sample image according to the color distribution diagram and the edge gray level distribution diagram.
3. An artificial intelligence based plant pest diagnosis method according to claim 1, wherein the method for obtaining disease style characteristics of each labeled plant sample image according to the color texture distribution map comprises:
processing the color texture distribution map by using a style migration algorithm to obtain a Gram matrix corresponding to each color texture distribution map;
and taking the Gram matrix as a disease style characteristic corresponding to each plant sample image with the label.
4. An artificial intelligence based plant pest diagnosis method as claimed in claim 1, wherein said method of training a variational self-encoder using said labeled plant sample images comprises:
and respectively inputting the plant sample images with the labels into a variation self-encoder, and training the variation self-encoder by utilizing a reconstruction loss function, an intra-class disease style loss function, an inter-class disease style loss function and a global loss function.
5. An artificial intelligence based plant pest diagnosis method according to claim 4 wherein said reconfiguration loss function is:
wherein,is as followsTagged second in individual status categoryThe variation corresponding to each plant sample image is from the reconstruction loss function of the encoder,is corresponding to plantTagged second in individual status categoryA Gaussian model corresponding to the image of each plant sample,is corresponding to plantThe gaussian model corresponding to each state class,is corresponding to plantTagged second in individual status categoryAn image of a sample of an individual plant,corresponding to plants output from the decoder of the encoder for variationTagged second in individual status categoryA reconstructed image corresponding to the image of the plant sample,is composed ofAndis/are as followsDivergence.
6. An artificial intelligence based plant pest diagnosis method according to claim 4 wherein said intra-category disease style loss function is:
wherein,is as followsTagged second in individual status categoryThe variation corresponding to each plant sample image is from the intra-category disease style loss function of the encoder,is corresponding to plantThe number of labeled plant specimen images in each status category,is corresponding to plantTagged second in individual status categoryAn image of a sample of an individual plant,is corresponding to plantTagged second in individual status categoryAn image of a sample of an individual plant,is corresponding to plantTagged second in individual status categoryThe plant sample image and the labeled secondThe degree of fusion of disease style characteristics of individual plant sample images,is as followsTagged second in individual status categoryThe plant sample image and the labeled secondGaussian model between plant sample imagesThe degree of fusion is such that,。
7. an artificial intelligence based plant pest diagnosis method according to claim 4 wherein said inter-class condition style loss function is:
wherein,is as followsTagged second in individual status categoryThe variation corresponding to each plant sample image is from the inter-class symptom style loss function of the encoder,the number of state categories corresponding to the plants,is corresponding to plantThe status of each of the plurality of status categories,is corresponding to plantThe status of each of the plurality of status categories,,is corresponding to plantStatus class andthe difficulty of gaussian model discrimination between individual state classes,is corresponding to plantStatus class andthe difficulty of distinguishing the disease style characteristics between individual status categories.
8. An artificial intelligence based plant pest diagnosis method according to claim 4 wherein the global loss function is:
wherein,is corresponding to plantTagged second in individual status categoryThe variation corresponding to each plant sample image is from the global loss function of the encoder,is corresponding to plantThe gaussian model corresponding to each state class,is a standard Gaussian model and is used as a model,is composed ofAndis/are as followsDivergence.
9. A plant disease and pest diagnosis method based on artificial intelligence as claimed in claim 1, wherein the method for obtaining label data corresponding to the plant sample image to be labeled comprises:
inputting each plant sample image to be labeled into a trained variational self-encoder to obtain a Gaussian model corresponding to each plant sample image to be labeled;
calculating the probability that each plant sample image to be labeled belongs to each state category corresponding to the plant according to the Gaussian model of each plant sample image to be labeled;
and taking the probability that each plant sample image to be labeled belongs to each state type corresponding to the plant as label data corresponding to the plant sample image to be labeled.
10. An artificial intelligence based plant pest diagnosis system comprising a memory and a processor, wherein the processor executes a computer program stored in the memory to implement an artificial intelligence based plant pest diagnosis method according to any one of claims 1 to 9.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111249226.1A CN113688959B (en) | 2021-10-26 | 2021-10-26 | Plant disease and insect pest diagnosis method and system based on artificial intelligence |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111249226.1A CN113688959B (en) | 2021-10-26 | 2021-10-26 | Plant disease and insect pest diagnosis method and system based on artificial intelligence |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113688959A true CN113688959A (en) | 2021-11-23 |
CN113688959B CN113688959B (en) | 2022-02-18 |
Family
ID=78588163
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111249226.1A Active CN113688959B (en) | 2021-10-26 | 2021-10-26 | Plant disease and insect pest diagnosis method and system based on artificial intelligence |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113688959B (en) |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102930249A (en) * | 2012-10-23 | 2013-02-13 | 四川农业大学 | Method for identifying and counting farmland pests based on colors and models |
JP2016168046A (en) * | 2015-03-09 | 2016-09-23 | 学校法人法政大学 | Plant disease diagnostic system, plant disease diagnostic method, and program |
CN106373136A (en) * | 2015-07-22 | 2017-02-01 | 崔胡晋 | Color and feature-based pest identifying and counting method |
CN109308697A (en) * | 2018-09-18 | 2019-02-05 | 安徽工业大学 | A kind of leaf disease recognition method based on machine learning algorithm |
CN110188824A (en) * | 2019-05-31 | 2019-08-30 | 重庆大学 | A kind of small sample plant disease recognition methods and system |
CN111507419A (en) * | 2020-04-22 | 2020-08-07 | 腾讯科技(深圳)有限公司 | Training method and device of image classification model |
CN112036335A (en) * | 2020-09-03 | 2020-12-04 | 南京农业大学 | Deconvolution-guided semi-supervised plant leaf disease identification and segmentation method |
CN112560964A (en) * | 2020-12-18 | 2021-03-26 | 深圳赛安特技术服务有限公司 | Method and system for training Chinese herbal medicine pest and disease identification model based on semi-supervised learning |
CN113468984A (en) * | 2021-06-16 | 2021-10-01 | 哈尔滨理工大学 | Crop pest and disease leaf identification system, identification method and pest and disease prevention method |
-
2021
- 2021-10-26 CN CN202111249226.1A patent/CN113688959B/en active Active
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102930249A (en) * | 2012-10-23 | 2013-02-13 | 四川农业大学 | Method for identifying and counting farmland pests based on colors and models |
JP2016168046A (en) * | 2015-03-09 | 2016-09-23 | 学校法人法政大学 | Plant disease diagnostic system, plant disease diagnostic method, and program |
CN106373136A (en) * | 2015-07-22 | 2017-02-01 | 崔胡晋 | Color and feature-based pest identifying and counting method |
CN109308697A (en) * | 2018-09-18 | 2019-02-05 | 安徽工业大学 | A kind of leaf disease recognition method based on machine learning algorithm |
CN110188824A (en) * | 2019-05-31 | 2019-08-30 | 重庆大学 | A kind of small sample plant disease recognition methods and system |
CN111507419A (en) * | 2020-04-22 | 2020-08-07 | 腾讯科技(深圳)有限公司 | Training method and device of image classification model |
CN112036335A (en) * | 2020-09-03 | 2020-12-04 | 南京农业大学 | Deconvolution-guided semi-supervised plant leaf disease identification and segmentation method |
CN112560964A (en) * | 2020-12-18 | 2021-03-26 | 深圳赛安特技术服务有限公司 | Method and system for training Chinese herbal medicine pest and disease identification model based on semi-supervised learning |
CN113468984A (en) * | 2021-06-16 | 2021-10-01 | 哈尔滨理工大学 | Crop pest and disease leaf identification system, identification method and pest and disease prevention method |
Non-Patent Citations (3)
Title |
---|
G. MADHULATHA: "Recognition of Plant Diseases using Convolutional Neural Network", 《2020 FOURTH INTERNATIONAL CONFERENCE ON I-SMAC (IOT IN SOCIAL, MOBILE, ANALYTICS AND CLOUD) (I-SMAC)》 * |
涂智潇: "基于计算机视觉的农作物病虫害检测系统", 《中国优秀硕士学位论文全文数据库 农业科技辑》 * |
蔡高勇: "基于深度卷积生成对抗网络的小样本植物病害识别问题研究", 《中国优秀硕士学位论文全文数据库 农业科技辑》 * |
Also Published As
Publication number | Publication date |
---|---|
CN113688959B (en) | 2022-02-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Aich et al. | Deepwheat: Estimating phenotypic traits from crop images with deep learning | |
CN111985536B (en) | Based on weak supervised learning gastroscopic pathology image Classification method | |
Al Bashish et al. | Detection and classification of leaf diseases using K-means-based segmentation and | |
CN114341948A (en) | System and method for plant species identification | |
CN109325431B (en) | Method and device for detecting vegetation coverage in feeding path of grassland grazing sheep | |
Ennadifi et al. | Wheat diseases classification and localization using convolutional neural networks and GradCAM visualization | |
Al-Badri et al. | Classification of weed using machine learning techniques: a review—challenges, current and future potential techniques | |
Liang et al. | Low-cost weed identification system using drones | |
Kirola et al. | Plants diseases prediction framework: A image-based system using deep learning | |
CN108629289A (en) | The recognition methods in farmland and system, applied to the unmanned plane of agricultural | |
CN115205833A (en) | Method and device for classifying growth states of cotton with few samples | |
Tosawadi et al. | On the use of class activation map on rice blast disease identification and localization | |
CN113688959B (en) | Plant disease and insect pest diagnosis method and system based on artificial intelligence | |
Balram et al. | Crop field monitoring and disease detection of plants in smart agriculture using internet of things | |
Joshi et al. | Detection and classification of plant diseases using soft computing techniques | |
CN115063602A (en) | Crop pest and disease identification method based on improved YOLOX-S network | |
CN115170987A (en) | Method for detecting diseases of grapes based on image segmentation and registration fusion | |
Cao et al. | Plant leaf segmentation and phenotypic analysis based on fully convolutional neural network | |
Attada et al. | A methodology for automatic detection and classification of pests using optimized SVM in greenhouse crops | |
Pajares et al. | Combining classifiers through fuzzy cognitive maps in natural images | |
Bharti et al. | Detection and classification of plant diseases | |
GSR et al. | COTTON LEAF DISEASE DETECTION USING ARTIFICIAL NEURAL NETWORK. | |
Schurischuster | Image analysis approaches for parasite detection on honeybees | |
CN118172676B (en) | Farmland pest detection method based on quantum deep learning | |
Chaudhari et al. | Identification of Banana Disease Using Color and Texture Feature |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |