Wolfberry disease and pest monitoring system and method based on image recognition

Abstract

The invention discloses a wolfberry disease and insect pest monitoring system and method based on image recognition, and belongs to the technical field of machine vision. The system comprises a synchronous data acquisition system, a first data processing system and a second data processing system, the synchronous data acquisition system is used for synchronously acquiring radar data and image data of a to-be-detected area; the first data processing system is used for performing feature analysis and region segmentation on the radar data to obtain a pest distribution region; performing pest type identification on the pest distribution area by using the image data to obtain a pest identification result; and the second data processing system is used for carrying out wolfberry disease identification analysis based on the image data to obtain a disease analysis result. According to the method, the radar data and the image data are processed and analyzed, so that Chinese wolfberry diseases and insect pests are identified.

Description

Technical Field

[0001] The present invention belongs to the technical field of machine vision, and in particular relates to a wolfberry disease and insect pest monitoring system and method based on image recognition. Background Art

[0002] As an important medicinal and economic crop, wolfberry is often attacked by a variety of diseases and pests during its growth, which seriously affects the yield and quality of wolfberry. Therefore, it is of great significance to accurately identify the types of wolfberry diseases and pests, analyze their symptoms and damage characteristics, understand their occurrence patterns, and explore effective prevention and control technologies to ensure the sustainable development of the wolfberry industry.

[0003] Traditional identification methods include visual observation, symptom comparison, biological indicator monitoring, and empirical judgment. Traditional identification methods for wolfberry pests and diseases have met the needs of traditional agricultural production to a certain extent, but with the advancement of agricultural modernization and the increasing number of pests and diseases, the limitations of these methods have become increasingly prominent. Visual observation and empirical judgment are highly subjective and their accuracy is difficult to guarantee; magnifying glass inspection and chemical reagent testing are complex and costly; symptom comparison and biological indicator monitoring rely on complete reference materials and long-term data accumulation. Therefore, in the prevention and control of wolfberry pests and diseases, it is necessary to continuously explore and apply new technologies and methods to improve the accuracy and efficiency of pest and disease identification in order to ensure the healthy development of the wolfberry industry. Summary of the invention

[0004] The present invention aims to solve the deficiencies of the prior art and proposes a wolfberry disease and insect pest monitoring system and method based on image recognition, which can realize the identification of wolfberry diseases and insect pests by processing and analyzing radar data and image data.

[0005] To achieve the above object, the present invention provides the following scheme: a wolfberry pest and disease monitoring system based on image recognition, comprising: a synchronous data acquisition system, a first data processing system and a second data processing system;

[0006] The synchronous data acquisition system is used to synchronously acquire radar data and image data of the area to be detected;

[0007] The first data processing system is used to perform feature analysis and region segmentation on the radar data to obtain pest distribution areas; and use the image data to identify pest types in the pest distribution areas to obtain pest identification results;

[0008] The second data processing system is used to perform wolfberry disease identification and analysis based on the image data to obtain a disease analysis result.

[0009] Further preferably, the first data processing system comprises: an analysis and processing module, a feature extraction module, an identification and segmentation module, and an insect pest identification module;

[0010] The analysis and processing module is used to perform feature analysis and image preprocessing on the radar data to obtain a processed image;

[0011] The feature extraction module is used to extract features from the processed image to obtain pest features;

[0012] The identification and segmentation module is used to segment the extracted pest features into pest areas to obtain pest areas;

[0013] The pest identification module is used to analyze and identify the image data based on the pest area to obtain the pest identification result.

[0014] Further preferably, the second data processing system comprises: an image preprocessing module, a leaf recognition module and a disease recognition module;

[0015] The image preprocessing module is used to perform image preprocessing on the image data to obtain a recognition image;

[0016] The leaf recognition module is used to perform leaf recognition on the recognition image and perform image segmentation on the recognized leaves to obtain leaf data;

[0017] The disease identification module is used to perform disease analysis on the blade data to obtain the disease analysis result.

[0018] Further preferably, the leaf recognition module comprises: a target recognition unit and an image segmentation unit;

[0019] The target recognition unit is used to construct a target recognition model, and perform leaf recognition on the recognition image based on the target recognition model to obtain a recognition result;

[0020] The image segmentation unit is used to separate the leaves based on the recognition result to obtain the leaf data.

[0021] The present invention also provides a method for monitoring wolfberry pests and diseases based on image recognition, comprising the following steps:

[0022] S1, synchronously acquiring radar data and image data of wolfberries in the detection area;

[0023] S2, performing feature analysis and region segmentation on the acquired radar data to obtain a pest distribution area; and using the image data to identify the pest type in the pest distribution area to obtain a pest identification result;

[0024] S3. Perform wolfberry disease identification analysis on the image data to obtain a disease analysis result.

[0025] Further preferably, S2 comprises the following steps:

[0026] S21, performing feature analysis and image preprocessing on the radar data to obtain a processed image;

[0027] S22, extracting features from the processed image to obtain pest features;

[0028] S23, performing pest area segmentation on the extracted pest features to obtain pest areas;

[0029] S24: Analyze and identify the image data based on the pest area to obtain the pest identification result.

[0030] Further preferably, S3 comprises the following steps:

[0031] S31, performing image preprocessing on the image data to obtain a recognition image;

[0032] S32, performing leaf recognition on the recognition image, and performing image segmentation on the recognized leaves to obtain leaf data;

[0033] S33, performing disease analysis on the leaf data to obtain the disease analysis result.

[0034] Further preferably, in S32, a target recognition model is constructed, and leaf recognition is performed on the recognition image based on the target recognition model to obtain a recognition result; and an image segmentation algorithm is used to separate the leaves of the recognition result to obtain the leaf data.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] The present invention realizes the identification of wolfberry diseases and pests by processing and analyzing radar data and image data. First, based on the characteristics of radar data, the pest area is determined, and then the pest type is obtained by analyzing the corresponding image data. At the same time, the image data is separated into leaves and fruits to obtain leaf data, and the leaf data is analyzed to obtain the disease type. The scheme of the present invention can reduce the calculation time and increase the processing efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0038] Figure 1The figure is a flow chart of a method for monitoring wolfberry pests and diseases based on image recognition according to an embodiment of the present invention. DETAILED DESCRIPTION

[0039] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0040] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0041] Embodiment 1:

[0042] This embodiment provides a wolfberry pest and disease monitoring system based on image recognition, including: a synchronous data acquisition system, a first data processing system and a second data processing system.

[0043] Among them, the synchronous data acquisition system is used to synchronously acquire radar data and image data of the area to be detected.

[0044] Relevant data information of the area to be inspected is collected synchronously through drones or other platforms equipped with laser radar equipment and image acquisition equipment such as cameras.

[0045] The first data processing system is used to perform feature analysis and region segmentation on radar data to obtain pest distribution areas; and use image data to identify pest types in the pest distribution areas to obtain pest identification results.

[0046] Specifically, the first data processing system includes: an analysis and processing module, a feature extraction module, an identification and segmentation module, and a pest identification module.

[0047] The analysis and processing module is used to perform feature analysis and image preprocessing on radar data to obtain a processed image.

[0048] For the acquired radar image, first, it is divided into several pixel blocks, and then the pixel blocks are further adaptively divided to obtain sub-blocks. Each sub-block has an imaging pixel node that uniquely corresponds to a calibration point of a regional feature. The Gaussian mixture model is used to describe the pixel labeling process. At the same time, the characteristic function expression of the area to be detected is obtained, which can effectively reflect the pest area information in the radar data.

[0049] In this embodiment, the characteristic function expression is:

[0050] T=1-ge -P,

[0051] Where g represents the probability density distribution function of the marking point; P represents the probability density function of the label field.

[0052] in,

[0053] Where K represents the total number of segmentation paths of radar data; k , μ k , σ k} is the set of uncertain parameters in the Gaussian mixture model; μ k , σ k 2 They represent the mean and variance of the k-th Gaussian distribution in the local neighborhood of the radar image respectively; x i Represents the pixels of the labeled field sub-block.

[0054]

[0055] Wherein, w3 represents the numbered field sub-block.

[0056] Afterwards, based on the feature analysis, the radar image is smoothed to obtain a processed image. The smoothing process includes:

[0057] A regional prior model of the radar image is constructed, and then a smoothing processing model is constructed based on the prior model, and the radar image is smoothed by the smoothing processing model.

[0058] The regional prior model is described by Gibbs distribution, specifically:

[0059]

[0060] Where β represents the global parameter of Gibbs distribution, which is used to control the smoothness; represents the local energy function.

[0061] Smoothing models include:

[0062]

[0063] in, Smoothing control parameters representing the area where pests occur; are the local control parameters in the four directions (horizontal, vertical, diagonal and anti-diagonal) of the radar data; k represents the number of marking points; y i k Represents the smoothing threshold of the i-th labeled node.

[0064] The feature extraction module is used to extract features from the processed image to obtain pest features.

[0065] The feature extraction process includes: segmenting the graph model structure in the horizontal, vertical and diagonal directions to obtain a statistical feature set; and obtaining a local distribution description of the pest area through the statistical set:

[0066] I(x)=J(x)t(x)+A(1-t(x)),

[0067] Where J(x) represents the global features of the radar image; t(x) represents the superpixel set; A represents the statistical feature set.

[0068] Calculate the positive definite tensor of the local area of ​​the radar image and obtain the constraint parameters of the local features:

[0069] h=θ / π,

[0070] s=1-λ2 / λ1,

[0071] v = λ1 + λ2;

[0072] Where θ represents the imaging tensor angle of the pest area; λ1 and λ2 are the long and short semi-axis lengths of the edge detection of the pest area, respectively.

[0073] Afterwards, according to the above parameters, the feature extraction results are obtained:

[0074]

[0075] Where δ and ε represent the convergence coefficient and insect corrosion operator respectively.

[0076] The recognition and segmentation module is used to segment the extracted pest features into pest areas to obtain pest areas.

[0077] The methods of region segmentation include:

[0078] Based on the above feature extraction results of the pest area, the probability distribution of accurate segmentation of the pest area is obtained, and the regional segmentation problem is converted into a problem of maximizing the likelihood distribution probability, thereby realizing the segmentation of the pest area.

[0079] The pest identification module is used to analyze and identify image data based on the pest area to obtain pest identification results.

[0080] The frame range of the pest coverage area is obtained through the above pest area segmentation results. After that, the synchronously collected image data is obtained, and pest recognition is performed on the image data to obtain pest recognition results. In this embodiment, pest recognition and the following disease recognition use the same pest recognition model, which uses the pest data set to train the target recognition model to obtain the recognition model of pest recognition. Then, the image data is analyzed and recognized to obtain the corresponding pest recognition results.

[0081] The method for constructing the pest dataset includes: selecting 2870 images to form the pest dataset based on the distance of the scene during shooting and the occlusion category of the occluded fruit. The distance is based on the distance of the shooting, and the distance within 1m is the close view. The light or heavy occlusion is determined by the visibility of the overall outline of the occluded leaf, and the overall outline is visible as light occlusion. After that, LabelImg is used to annotate the captured images, and image data enhancement technology is introduced to expand the image dataset by adjusting image brightness, hue, noise, flipping, translation, cropping, rotation and other operations.

[0082] The pest and disease recognition model uses an improved YOLOv7 model to identify pests and diseases. The model mainly includes: input layer, backbone network, neck and prediction layer. Among them, the convolution layer in the backbone network and the neck is replaced by a phantom convolution layer, the standard convolution is replaced by a linear transformation, and the redundant features are replaced by simple features, which reduces the complexity of the model. And add a CBAM attention module to the backbone network. By combining the channel attention module and the spatial attention model, the CBAM attention module can use the spatial attention mechanism to make the neural network focus on the classification area that plays a decisive role in the feature image, and can also use the channel attention mechanism to process the distribution relationship of the feature map channel, so that the neural network can perform adaptive feature optimization along these two dimensions. Among them, the channel attention module includes maximum pooling and average pooling to transform the size of the input feature map; then it is converted by a multi-layer perceptron, and then it is summed and normalized to obtain the channel attention feature map. The spatial attention module performs global average pooling and global maximum pooling operations on the input channel attention feature map to obtain two spatial feature maps, and then performs convolution operations on them to generate the spatial attention feature map. Then, the feature map of the original input and the spatial attention feature map are multiplied element by element to obtain the feature map of the attention module. At the same time, a weighted bidirectional feature pyramid network is added in the neck to better retain the feature information of small objects.

[0083] The second data processing system is used to perform wolfberry disease identification and analysis based on image data to obtain disease analysis results. Among them, the second data processing system includes: an image preprocessing module, a leaf recognition module and a disease recognition module; the image preprocessing module is used to perform image preprocessing on the image data to obtain a recognition image; the leaf recognition module is used to perform leaf recognition on the recognition image, and perform image segmentation on the recognized leaves to obtain leaf data; the disease recognition module is used to perform disease analysis on the leaf data to obtain disease analysis results. Further, the leaf recognition module includes: a target recognition unit and an image segmentation unit; the target recognition unit is used to construct a target recognition model, and perform leaf recognition on the recognition image based on the target recognition model to obtain a recognition result; the image segmentation unit is used to separate the leaves based on the recognition result to obtain leaf data.

[0084] In this embodiment, the method of image preprocessing includes: removing useless environmental background interference noise by limiting the distance between the foreground and background depths; and performing normalization, contrast enhancement, and data expansion operations at the same time. Afterwards, the image data size is unified to facilitate improving the network operation efficiency and recognition efficiency in the subsequent recognition and analysis process.

[0085] The target recognition model includes: a feature extraction module, an attention mechanism weighting module, a pooling layer, a global average pooling layer and a classification module. The feature extraction module adopts a high-efficiency network structure, which can adaptively adjust the relationship between the width, depth and image resolution in the network during the training process. Afterwards, the output of the feature extraction module is used as the input of the attention mechanism weighting module, and the obtained features are subjected to attention weighting operation. Specifically, the input features are operated by three convolution kernels, and converted into nonlinear features with stronger expression by activation functions Relu and Sigmoid. The output features of the feature extraction module and the nonlinear features output by the attention mechanism weighting module are multiplied to obtain an image mask. Afterwards, in order to avoid overfitting of the network, the generated image mask and nonlinear features are passed through the pooling layer. Finally, more feature information is obtained through the global average pooling layer, and the recognition target is screened out using the Softmax function. Afterwards, the threshold segmentation method is used for the screened leaf targets to obtain leaf data.

[0086] Embodiment 2:

[0087] like Figure 1 As shown, this embodiment provides a wolfberry pest and disease monitoring method based on image recognition, comprising the following steps:

[0088] S1. Synchronously obtain radar data and image data for the wolfberries in the detection area.

[0089] S2. Perform feature analysis and region segmentation on the acquired radar data to obtain the pest distribution area; and use image data to identify the pest type in the pest distribution area to obtain the pest identification result.

[0090] Further, S2 comprises the following steps:

[0091] S21, performing feature analysis and image preprocessing on the radar data to obtain a processed image. The specific method includes:

[0092] For the acquired radar image, first, it is divided into several pixel blocks, and then the pixel blocks are further adaptively divided to obtain sub-blocks. Each sub-block has an imaging pixel node that uniquely corresponds to a calibration point of a regional feature. The Gaussian mixture model is used to describe the pixel labeling process. At the same time, the characteristic function expression of the area to be detected is obtained, which can effectively reflect the pest area information in the radar data.

[0093] In this embodiment, the characteristic function expression is:

[0094] T=1-ge -P ,

[0095] Where g represents the probability density distribution function of the marking point; P represents the probability density function of the label field.

[0096] in,

[0097] Where K represents the total number of segmentation paths of radar data; k , μ k , σ k} is the set of uncertain parameters in the Gaussian mixture model; μ k , σ k 2 They represent the mean and variance of the k-th Gaussian distribution in the local neighborhood of the radar image respectively; x i Represents the pixels of the labeled field sub-block.

[0098]

[0099] Wherein, w3 represents the numbered field sub-block.

[0100] Afterwards, based on the feature analysis, the radar image is smoothed to obtain a processed image. The smoothing process includes:

[0101] A regional prior model of the radar image is constructed, and then a smoothing processing model is constructed based on the prior model, and the radar image is smoothed by the smoothing processing model.

[0102] The regional prior model is described by Gibbs distribution, specifically:

[0103]

[0104] Where β represents the global parameter of Gibbs distribution, which is used to control the smoothness; represents the local energy function.

[0105] Smoothing models include:

[0106]

[0107] Among them, β i =[β i 1 , β i 2 , β i 3 , β i 4 ] T represents the smoothing control parameter of the pest area; β i 1 , β i 2 , β i 3 , β i 4 are the local control parameters in the four directions (horizontal, vertical, diagonal and anti-diagonal) of the radar data; k represents the number of marking points; y i k Represents the smoothing threshold of the i-th labeled node.

[0108] S22, extracting features from the processed image to obtain pest features.

[0109] The feature extraction process includes: segmenting the graph model structure in the horizontal, vertical and diagonal directions to obtain a statistical feature set; and obtaining a local distribution description of the pest area through the statistical set:

[0110] I(x)=J(x)t(x)+A(1-t(x)),

[0111] Where J(x) represents the global features of the radar image; t(x) represents the superpixel set; A represents the statistical feature set.

[0112] Calculate the positive definite tensor of the local area of ​​the radar image and obtain the constraint parameters of the local features:

[0113] h=θ / π,

[0114] s=1-λ2 / λ1,

[0115] v = λ1 + λ2;

[0116] Where θ represents the imaging tensor angle of the pest area; λ1 and λ2 are the long and short semi-axis lengths of the edge detection of the pest area, respectively.

[0117] Afterwards, according to the above parameters, the feature extraction results are obtained:

[0118]

[0119] Where δ and ε represent the convergence coefficient and insect corrosion operator respectively.

[0120] S23, performing pest area segmentation on the extracted pest features to obtain the pest area.

[0121] The methods of region segmentation include:

[0122] Based on the above feature extraction results of the pest area, the probability distribution of accurate segmentation of the pest area is obtained, and the regional segmentation problem is converted into a problem of maximizing the likelihood distribution probability, thereby realizing the segmentation of the pest area.

[0123] S24: Analyze and identify the image data based on the pest area to obtain a pest identification result.

[0124] The frame range of the pest coverage area is obtained through the above pest area segmentation results. After that, the synchronously collected image data is obtained, and pest recognition is performed on the image data to obtain pest recognition results. In this embodiment, pest recognition and the following disease recognition use the same pest recognition model, which uses the pest data set to train the target recognition model to obtain the recognition model of pest recognition. Then, the image data is analyzed and recognized to obtain the corresponding pest recognition results.

[0125] The pest and disease recognition model uses an improved YOLOv7 model to identify pests and diseases. The model mainly includes: input layer, backbone network, neck and prediction layer. Among them, the convolution layer in the backbone network and the neck is replaced by a phantom convolution layer, the standard convolution is replaced by a linear transformation, and the redundant features are replaced by simple features, which reduces the complexity of the model. And add a CBAM attention module to the backbone network. By combining the channel attention module and the spatial attention model, the CBAM attention module can use the spatial attention mechanism to make the neural network focus on the classification area that plays a decisive role in the feature image, and can also use the channel attention mechanism to process the distribution relationship of the feature map channel, so that the neural network can perform adaptive feature optimization along these two dimensions. Among them, the channel attention module includes maximum pooling and average pooling to transform the size of the input feature map; then it is converted by a multi-layer perceptron, and then it is summed and normalized to obtain the channel attention feature map. The spatial attention module performs global average pooling and global maximum pooling operations on the input channel attention feature map to obtain two spatial feature maps, and then performs convolution operations on them to generate the spatial attention feature map. Then, the feature map of the original input and the spatial attention feature map are multiplied element by element to obtain the feature map of the attention module. At the same time, a weighted bidirectional feature pyramid network is added in the neck to better retain the feature information of small objects.

[0126] S3. Perform wolfberry disease recognition analysis on the image data to obtain disease analysis results.

[0127] Further, S3 includes the following steps:

[0128] S31, performing image preprocessing on the image data to obtain a recognition image.

[0129] The image preprocessing method includes: removing useless environmental background interference noise by limiting the distance between the foreground and background depth; performing normalization, contrast enhancement and data expansion operations at the same time. After that, the image data size is unified to improve the network operation efficiency and recognition efficiency in the subsequent recognition and analysis process.

[0130] S32, performing leaf recognition on the recognition image, and performing image segmentation on the recognized leaves to obtain leaf data; obtaining a recognition result by constructing a target recognition model, and performing leaf recognition on the recognition image based on the target recognition model; and using an image segmentation algorithm to separate the leaves from the recognition result to obtain leaf data.

[0131] The target recognition model includes: a feature extraction module, an attention mechanism weighting module, a pooling layer, a global average pooling layer and a classification module. The feature extraction module adopts a high-efficiency network structure, which can adaptively adjust the relationship between the width, depth and image resolution in the network during the training process. Afterwards, the output of the feature extraction module is used as the input of the attention mechanism weighting module, and the obtained features are subjected to attention weighting operation. Specifically, the input features are operated by three convolution kernels, and converted into nonlinear features with stronger expression by activation functions Relu and Sigmoid. The output features of the feature extraction module and the nonlinear features output by the attention mechanism weighting module are multiplied to obtain an image mask. Afterwards, in order to avoid overfitting of the network, the generated image mask and nonlinear features are passed through the pooling layer. Finally, more feature information is obtained through the global average pooling layer, and the recognition target is screened out using the Softmax function. Afterwards, the threshold segmentation method is used for the screened leaf targets to obtain leaf data.

[0132] S33, performing disease analysis on the leaf data to obtain disease analysis results.

[0133] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. The wolfberry pest and disease monitoring system based on image recognition is characterized by: include: Synchronizing the data acquisition system, the first data processing system, and the second data processing system; The synchronous data acquisition system is used to synchronously acquire radar data and image data of the area to be detected; The first data processing system is used to perform feature analysis and region segmentation on the radar data to obtain pest distribution areas; and use the image data to identify pest types in the pest distribution areas to obtain pest identification results; The second data processing system is used to perform wolfberry disease identification and analysis based on the image data to obtain a disease analysis result.

2. The wolfberry pest monitoring system based on image recognition according to claim 1 is characterized in that: The first data processing system includes: an analysis and processing module, a feature extraction module, an identification and segmentation module, and an insect pest identification module; The analysis and processing module is used to perform feature analysis and image preprocessing on the radar data to obtain a processed image; The feature extraction module is used to extract features from the processed image to obtain pest features; The identification and segmentation module is used to segment the extracted pest features into pest areas to obtain pest areas; The pest identification module is used to analyze and identify the image data based on the pest area to obtain the pest identification result.

3. The wolfberry pest monitoring system based on image recognition according to claim 1 is characterized in that: The second data processing system includes: an image preprocessing module, a leaf recognition module and a disease recognition module; The image preprocessing module is used to perform image preprocessing on the image data to obtain a recognition image; The leaf recognition module is used to perform leaf recognition on the recognition image and perform image segmentation on the recognized leaves to obtain leaf data; The disease identification module is used to perform disease analysis on the blade data to obtain the disease analysis result.

4. The wolfberry pest monitoring system based on image recognition according to claim 3 is characterized in that: The leaf recognition module includes: a target recognition unit and an image segmentation unit; The target recognition unit is used to construct a target recognition model, and perform leaf recognition on the recognition image based on the target recognition model to obtain a recognition result; The image segmentation unit is used to separate the leaves based on the recognition result to obtain the leaf data.

5. A method for monitoring wolfberry pests and diseases based on image recognition, the method being applied to the system according to any one of claims 1 to 4, characterized in that: The following steps are involved: S1, synchronously acquiring radar data and image data of wolfberries in the detection area; S2, performing feature analysis and region segmentation on the acquired radar data to obtain a pest distribution area; and using the image data to identify the pest type in the pest distribution area to obtain a pest identification result; S3. Perform wolfberry disease identification analysis on the image data to obtain a disease analysis result.

6. The method for monitoring wolfberry pests and diseases based on image recognition according to claim 5, characterized in that: S2 includes the following steps: S21, performing feature analysis and image preprocessing on the radar data to obtain a processed image; S22, extracting features from the processed image to obtain pest features; S23, performing pest area segmentation on the extracted pest features to obtain pest areas; S24: Analyze and identify the image data based on the pest area to obtain the pest identification result.

7. The method for monitoring wolfberry pests and diseases based on image recognition according to claim 5, characterized in that: S3 includes the following steps: S31, performing image preprocessing on the image data to obtain a recognition image; S32, performing leaf recognition on the recognition image, and performing image segmentation on the recognized leaves to obtain leaf data; S33, performing disease analysis on the leaf data to obtain the disease analysis result.

8. The method for monitoring wolfberry pests and diseases based on image recognition according to claim 7, characterized in that: In S32, a target recognition model is constructed, and leaf recognition is performed on the recognition image based on the target recognition model to obtain a recognition result; and an image segmentation algorithm is used to separate the leaves from the recognition result to obtain the leaf data.

Images