Pest and disease detection system and detection method for agricultural planting

Abstract

The invention discloses a disease and insect pest detection system for agricultural planting, and relates to the technical field of image analysis, the disease and insect pest detection system comprises a scanning module, an information processing module, a leaf segmentation module and a disease and insect pest qualitative module, the information processing module carries out space-time synchronization on original data from the visible light camera, the infrared thermal camera and the optical radar and carries out preprocessing on the original data, and the leaf segmentation module is responsible for inputting the processed data into a proper segmentation model so as to realize accurate identification and segmentation of crops in a planting area. The plant disease and insect pest qualitative module is responsible for carrying out qualitative treatment on plant disease and insect pest conditions according to leaf segmentation results, the scanning module comprises a visible light camera, an infrared thermal camera and an optical radar, the visible light camera is used for capturing high-resolution images above legume crops in a planting area, and the plant disease and insect pest qualitative detection system has the advantage of omnibearing detection.

Description

Technical Field

[0001] This invention relates to the field of image analysis technology, specifically to an agricultural pest and disease detection system and method. Background Technology

[0002] In agriculture, many plants may be threatened by multiple pests and diseases simultaneously, especially some legume crops with multi-layered canopies. Some pests, such as caterpillars, eat away at the leaves, causing leaf damage; others, such as aphids, suck sap, leading to insufficient water in the leaves and affecting the plant's health; and some parasitic viruses can infect plant cells, causing leaves to curl.

[0003] Existing pest and disease detection technologies rely on drones equipped with cameras to take pictures of plants and analyze the images to detect pests and diseases. However, this method usually only provides two-dimensional images, making it difficult to effectively distinguish the state changes between overlapping leaves. It is less effective for legumes with multi-layered canopies and cannot detect leaf water content. Therefore, it is necessary to design a comprehensive pest and disease detection system and method for agricultural planting. Summary of the Invention

[0004] The purpose of this invention is to provide an agricultural pest and disease detection system and method to solve the problems mentioned in the background art.

[0005] To address the aforementioned technical problems, this invention provides the following technical solution: an agricultural pest and disease detection system, comprising a scanning module, an information processing module, a leaf segmentation module, and a pest and disease characterization module. The scanning module is responsible for collecting multimodal data under different environmental conditions. The information processing module performs spatiotemporal synchronization of raw data from visible light cameras, infrared thermal cameras, and optical radar, and preprocesses the raw data. The leaf segmentation module is responsible for inputting the processed data into an appropriate segmentation model to achieve accurate identification and segmentation of crops in the planting area. The pest and disease characterization module is responsible for characterizing the pest and disease situation based on the results of the leaf segmentation module.

[0006] According to the above technical solution, the scanning module includes a visible light camera, an infrared thermal camera, and an optical radar. The visible light camera is used to capture high-resolution images of the legumes above the planting area and identify the characteristics of the legumes in the planting area, including color, defects, and shape. The infrared thermal camera is used to detect the infrared spectral characteristics of the planting area. The optical radar obtains distance information and spatial positioning data of the planting area through laser scanning, thereby constructing a three-dimensional model of the legumes in the planting area. The leaf segmentation module includes a visual leaf segmentation module, a point cloud leaf segmentation module, and a multimodal combination module. The visual leaf segmentation module performs pixel-level segmentation of legume crop leaf images from visible light and infrared images. The point cloud leaf segmentation module is used to perform fine grouping of point cloud data. The multimodal combination module is used to combine the results of the visual leaf segmentation module and the point cloud leaf segmentation module.

[0007] This invention also provides a method for detecting agricultural pests and diseases, specifically including the following steps: S1. After the drone is started, the visible light camera, infrared thermal camera and optical radar in the scanning module start working, respectively collecting images of the planting area, infrared spectral characteristics and three-dimensional spatial raw data. These raw data will be transmitted to the information processing module in real time. S2. The information processing module first performs spatiotemporal synchronization on the raw data from the visible light camera, infrared thermal camera, and LiDAR. Using the visible light camera as a reference, the coordinate systems of the infrared thermal camera and LiDAR are projected onto the coordinate system of the visible light camera to ensure the consistency of all data in time and space. The module then processes the data to produce images of legume leaves. Next, the data from the visible light and infrared thermal cameras are combined in a multimodal manner to generate a composite image. For the LiDAR data, the statistical outlier removal method is used to remove noise and outliers from the point cloud data. The processed image data and point cloud data are then transmitted to the corresponding leaf segmentation modules. S3. The Mask-RCNN model is used to segment the combined visible light and infrared images of legume leaves, generating image segmentation masks for legume leaves. The PV-RCNN model is used to perform target detection on the point cloud data of the optical radar, and the point cloud is finely grouped by the clustering algorithm to generate point cloud segmentation masks for legume leaves. Then, the two leaf segmentation masks are combined in a multimodal manner. For segmentation masks with overlap, the result with high confidence is selected as the final segmentation mask, while segmentation masks without overlap are directly merged. The two segmentation information are integrated to generate the final crop detection result. S4. Based on the leaf defect information, leaf curling information, and leaf water content information of legume crops provided by the leaf segmentation module, the severity of pests and diseases in legume crops is analyzed. Under normal cultivation conditions, the more severe the leaf defects detected in the leaf image, the more severe the damage from caterpillars; the higher the absorption of the near-infrared band in the leaf image, the more abundant the water content, and the less severe the aphid damage; and the higher the degree of curling in the leaf image, the more severe the damage from parasitic viruses.

[0008] According to the above technical solution, in step S2, when performing multimodal combination of data from visible light and infrared thermal imaging cameras, the specific steps are as follows: S21. Use a bilateral filter to denoise both images while preserving edge information; S22. Perform color correction on the visible light image using adaptive histogram equalization to improve image contrast, and perform brightness matching on the infrared image, using histogram matching to align its brightness distribution with the visible light image. S23. The visible light and infrared images are decomposed into a visual layer containing high-frequency information and a thermal imaging layer containing low-frequency information by using the Laplace pyramid decomposition method, and their respective image features are extracted. S24. In the weighted combination stage, the visual layer and the thermal imaging layer are combined separately. The high-frequency information of the visible light image contains edges and details, showing the outline and defect features of the leaves of legume crops, while the low-frequency information of the infrared image contains the distribution of heat sources, reflecting the infrared spectral characteristics of the leaves of legume crops, which are represented as smooth and significant areas.

[0009] According to the above technical solution, the combination formula of the visual layer and the thermal imaging layer in S24 is as follows: ; in, For position Superimposed brightness at the location, For visible light images at position Brightness at that location For infrared images at location The brightness at each location is calculated, and the two values ​​are then superimposed on the combined image. and For adaptive dynamic weights, the specific weights are determined by the following formula: , ; In the above formula, The degree of abrupt change in the visible light image. The reflectance is in the near-infrared band, and the degree of abrupt change in the visible light image is determined by position. The rate of change of brightness and color at a location with spatial position A high degree of abrupt change in the visible light image indicates a defect in the leaf at that location. Higher reflectivity in the near-infrared band indicates lower water content in the leaf, thus highlighting the defective and low-water-content areas more prominently.

[0010] According to the above technical solution, the multimodal combination process in S3 is specifically as follows: S31. When determining the confidence level of a point cloud segmentation mask, it is necessary to evaluate the density of points within a cluster. Number of clusters of the same category To determine, if the density of points within a cluster... The higher the value, the more complex the spatial morphology of the clusters, the more pronounced the variation in vertical distance between leaves, the greater the curl of the leaves, and the higher the number of clusters in the same category. The larger the value, the more prevalent this type of cluster is in the current planting area, and therefore the higher the confidence level. For image segmentation masks, the confidence level is determined by the variance of the superimposed pixel values ​​at all locations within the legume leaf image. The larger the variance, the more severe the leaf damage, and the higher the confidence level.

[0011] According to the above technical solution, the specific calculation method for the confidence level in step S31 is as follows: Confidence of point cloud segmentation mask ,in This is the density weighting coefficient. This is the weighting coefficient for the number of clusters. This is a conversion factor for the planting area. The planting area is the total area of ​​the planting area. The larger the area, the more likely it is to form clusters of the same type, so it is necessary to remove interference from the properties of the planting area itself; Confidence of image segmentation mask ,in This represents the number of pixels in an image of a legume leaf. This is the index of a specific pixel. The average superimposed pixel value of all pixels in the image segmentation mask. Since there is no need to determine whether there are problems such as low water content and defects when the leaf is severely curled, when... Then the point cloud segmentation mask is selected as the final segmentation mask. Then, the image segmentation mask is selected as the final segmentation mask. This is the conversion factor for variance when comparing confidence levels.

[0012] According to the above technical solution, in step S31, in order to reduce the confidence level of the image segmentation mask... The computational complexity during the calculation involves dividing the region of an image segmentation mask into a cross-shaped grid, with the intersection of each grid line serving as the basis for calculating the confidence score. The pixels used in the process are substituted into the formula for calculating the variance. The calculation formula is obtained Approximate value, grid side length of cross grid lines .

[0013] Compared with the prior art, the beneficial effects achieved by the present invention are: This invention utilizes a drone equipped with a visible light camera and an infrared thermal camera, combined with optical radar, to comprehensively detect legumes with multi-layered canopy leaves and generate composite images. These composite images are then used to comprehensively assess damage from caterpillars, aphids, and parasitic viruses. The display weights of the visual images and infrared thermal imaging are dynamically adjusted, allowing the composite images to emphasize leaf damage and low water content. When radar point clouds overlap with the image, areas with severe leaf curling are prioritized for radar point cloud display. This allows for the highlighting of key information within a single image, facilitating a more intuitive display of pest and disease severity and improving assessment efficiency. Attached Figure Description

[0014] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a schematic diagram of the overall modular structure of the present invention. Detailed Implementation

[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] Please see Figure 1 This invention provides an agricultural pest and disease detection system, including a scanning module, an information processing module, a leaf segmentation module, and a pest and disease identification module. The scanning module is responsible for collecting multimodal data under different environmental conditions. The information processing module performs spatiotemporal synchronization on the raw data from visible light cameras, infrared thermal cameras, and optical radar, and preprocesses the raw data. The leaf segmentation module is responsible for inputting the processed data into the corresponding segmentation model to achieve accurate identification and segmentation of crops in the planting area. The pest and disease identification module is responsible for identifying the pest and disease situation based on the results of the leaf segmentation module.

[0017] The scanning module includes a visible light camera, an infrared thermal camera, and an optical radar. The visible light camera is used to capture high-resolution images of the legumes above the planting area and identify the characteristics of the legumes in the planting area, including color, defects, and shape. The infrared thermal camera is used to detect the infrared spectral characteristics of the planting area. The optical radar obtains distance information and spatial positioning data of the planting area through laser scanning, thereby constructing a three-dimensional model of the legumes in the planting area. The leaf segmentation module includes a visual leaf segmentation module, a point cloud leaf segmentation module, and a multimodal combination module. The visual leaf segmentation module performs pixel-level segmentation of legume crop leaf images from visible light and infrared images. The point cloud leaf segmentation module is used to perform fine grouping of point cloud data. The multimodal combination module is used to combine the results of the visual leaf segmentation module and the point cloud leaf segmentation module.

[0018] This invention also provides a method for detecting agricultural pests and diseases, specifically including the following steps: S1. After the drone is started, the visible light camera, infrared thermal camera and optical radar in the scanning module start working, respectively collecting images of the planting area, infrared spectral characteristics and three-dimensional spatial raw data. These raw data will be transmitted to the information processing module in real time. S2. The information processing module first performs spatiotemporal synchronization on the raw data from the visible light camera, infrared thermal camera, and LiDAR. Using the visible light camera as a reference, the coordinate systems of the infrared thermal camera and LiDAR are projected onto the coordinate system of the visible light camera to ensure the consistency of all data in time and space. The module then processes the data to produce images of legume leaves. Next, the data from the visible light and infrared thermal cameras are combined in a multimodal manner to generate a composite image. For the LiDAR data, the statistical outlier removal method is used to remove noise and outliers from the point cloud data. The processed image data and point cloud data are then transmitted to the corresponding leaf segmentation modules. S3. The Mask-RCNN model is used to segment the combined visible light and infrared images of legume leaves, generating image segmentation masks for legume leaves. The PV-RCNN model is used to perform target detection on the point cloud data of the optical radar, and the point cloud is finely grouped by the clustering algorithm to generate point cloud segmentation masks for legume leaves. Then, the two leaf segmentation masks are combined in a multimodal manner. For segmentation masks with overlap, the result with high confidence is selected as the final segmentation mask, while segmentation masks without overlap are directly merged. The two segmentation information are integrated to generate the final crop detection result. S4. Based on the leaf defect information, leaf curling information, and leaf water content information of legume crops provided by the leaf segmentation module, the severity of pests and diseases in legume crops is analyzed. Under normal cultivation conditions, the more severe the leaf defects detected in the leaf image, the more severe the damage from caterpillars; the higher the absorption of the near-infrared band in the leaf image, the more abundant the water content, and the less severe the aphid damage; and the higher the degree of curling in the leaf image, the more severe the damage from parasitic viruses.

[0019] In S2, when combining data from visible light and infrared thermal cameras in a multimodal manner, the specific steps are as follows: S21. Use a bilateral filter to denoise both images while preserving edge information; S22. Perform color correction on the visible light image using adaptive histogram equalization to improve image contrast, and perform brightness matching on the infrared image, using histogram matching to align its brightness distribution with the visible light image. S23. The visible light and infrared images are decomposed into a visual layer containing high-frequency information and a thermal imaging layer containing low-frequency information by using the Laplace pyramid decomposition method, and their respective image features are extracted. S24. In the weighted combination stage, the visual layer and the thermal imaging layer are combined separately. The high-frequency information of the visible light image contains edges and details, showing the outline and defect features of the leaves of legume crops, while the low-frequency information of the infrared image contains the distribution of heat sources, reflecting the infrared spectral characteristics of the leaves of legume crops, and is represented as a smooth and significant area. In S24, the combination formula for the visual layer and the thermal imaging layer is: ; in, For position Superimposed brightness at the location, For visible light images at position Brightness at that location For infrared images at location The brightness at each location is calculated, and the two values ​​are then superimposed on the combined image. and For adaptive dynamic weights, the specific weights are determined by the following formula: ,

[0020] In the above formula, The degree of abrupt change in the visible light image. The reflectance is in the near-infrared band, and the degree of abrupt change in the visible light image is determined by position. The rate of change of brightness and color at a location with spatial position A high degree of abrupt change in the visible light image indicates a defect in the leaf at that location. Higher reflectivity in the near-infrared band indicates lower water content in the leaf, thus highlighting the defective and low-water-content areas more prominently.

[0021] In S3, the multimodal combination process is as follows: S31. When determining the confidence level of a point cloud segmentation mask, it is necessary to evaluate the density of points within a cluster. Number of clusters of the same category To determine, if the density of points within a cluster... The higher the value, the more complex the spatial morphology of the clusters, the more pronounced the variation in vertical distance between leaves, the greater the curl of the leaves, and the higher the number of clusters in the same category. The larger the value, the more prevalent this type of cluster is in the current planting area, and therefore the higher the confidence level. For image segmentation masks, the confidence level is determined by the variance of the superimposed pixel values ​​at all locations within the legume leaf image. The larger the variance, the more severe the leaf damage, and the higher the confidence level. In S31, the confidence level is calculated as follows: Confidence of point cloud segmentation mask ,in This is the density weighting coefficient. This is the weighting coefficient for the number of clusters. This is a conversion factor for the planting area. The planting area is the total area of ​​the planting area. The larger the area, the more likely it is to form clusters of the same type, so it is necessary to remove interference from the properties of the planting area itself; Confidence of image segmentation mask ,in This represents the number of pixels in an image of a legume leaf. This is the index of a specific pixel. The average superimposed pixel value of all pixels in the image segmentation mask. Since there is no need to determine whether there are problems such as low water content and defects when the leaf is severely curled, when... Then the point cloud segmentation mask is selected as the final segmentation mask. Then, the image segmentation mask is selected as the final segmentation mask. This is the conversion factor for variance when comparing confidence levels; In S31, in order to reduce the confidence of the image segmentation mask... The computational complexity during the calculation involves dividing the region of an image segmentation mask into a cross-shaped grid, with the intersection of each grid line serving as the basis for calculating the confidence score. The pixels used in the process are substituted into the formula for calculating the variance. The calculation formula is obtained Approximate value, grid side length of cross grid lines .

[0022] because The larger the area, the more obvious the clustering characteristics of the planting area are under the LiDAR scan, the greater the degree of leaf curling, and the more complex the image. Therefore, more pixels need to be calculated to obtain visual and infrared image features; otherwise, the information collection will be incomplete.

[0023] By using drones equipped with visible light and infrared thermal cameras, combined with optical radar, comprehensive detection of multi-layered canopy leaves in legume crops is conducted, generating composite images. These composite images are then used to comprehensively assess damage from caterpillars, aphids, and parasitic viruses. The display weights of visual images and infrared thermal imaging are dynamically adjusted so that the composite images highlight leaf damage and low water content. When radar point clouds overlap with the images, areas with severe leaf curling are prioritized for radar point cloud display. This allows for the highlighting of key information on a single image, facilitating a more intuitive display of the severity of pests and diseases and improving assessment efficiency.

[0024] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0025] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A pest and disease detection system for agricultural planting, characterized in that: The system includes a scanning module, an information processing module, a leaf segmentation module, and a pest and disease identification module. The scanning module is responsible for collecting multimodal data under different environmental conditions. The information processing module performs spatiotemporal synchronization on the raw data from visible light cameras, infrared thermal cameras, and optical radar, and preprocesses the raw data. The leaf segmentation module is responsible for inputting the processed data into the corresponding segmentation model to achieve accurate identification and segmentation of crops in the planting area. The pest and disease identification module is responsible for identifying the pest and disease situation based on the results of the leaf segmentation module.

2. The agricultural pest and disease detection system according to claim 1, characterized in that: The scanning module includes a visible light camera, an infrared thermal camera, and an optical radar. The visible light camera is used to capture high-resolution images of the legumes above the planting area and identify the characteristics of the legumes in the planting area, including color, defects, and shape. The infrared thermal camera is used to detect the infrared spectral characteristics of the planting area. The optical radar obtains distance information and spatial positioning data of the planting area through laser scanning, thereby constructing a three-dimensional model of the legumes in the planting area. The leaf segmentation module includes a visual leaf segmentation module, a point cloud leaf segmentation module, and a multimodal combination module. The visual leaf segmentation module performs pixel-level segmentation of legume crop leaf images from visible light and infrared images. The point cloud leaf segmentation module is used to perform fine grouping of point cloud data. The multimodal combination module is used to combine the results of the visual leaf segmentation module and the point cloud leaf segmentation module.

3. A method for detecting pests and diseases in agricultural planting, characterized in that: The system for detecting agricultural pests and diseases as described in any one of claims 1-2 includes the following steps: S1. After the drone is started, the visible light camera, infrared thermal camera and optical radar in the scanning module start working, respectively collecting images of the planting area, infrared spectral characteristics and three-dimensional spatial raw data. These raw data will be transmitted to the information processing module in real time. S2. The information processing module first performs spatiotemporal synchronization on the raw data from the visible light camera, infrared thermal camera, and LiDAR. Using the visible light camera as a reference, the coordinate systems of the infrared thermal camera and LiDAR are projected onto the coordinate system of the visible light camera to ensure the consistency of all data in time and space. The module then processes the data to produce images of legume leaves. Next, the data from the visible light and infrared thermal cameras are combined in a multimodal manner to generate a composite image. For the LiDAR data, the statistical outlier removal method is used to remove noise and outliers from the point cloud data. The processed image data and point cloud data are then transmitted to the corresponding leaf segmentation modules. S3. The Mask-RCNN model is used to segment the combined visible light and infrared images of legume leaves, generating image segmentation masks for legume leaves. The PV-RCNN model is used to perform target detection on the point cloud data of the optical radar, and the point cloud is finely grouped by the clustering algorithm to generate point cloud segmentation masks for legume leaves. Then, the two leaf segmentation masks are combined in a multimodal manner. For segmentation masks with overlap, the result with high confidence is selected as the final segmentation mask, while segmentation masks without overlap are directly merged. The two segmentation information are integrated to generate the final crop detection result. S4. Based on the leaf defect information, leaf curling information, and leaf water content information of legume crops provided by the leaf segmentation module, the severity of pests and diseases in legume crops is analyzed. Under normal cultivation conditions, the more severe the leaf defects detected in the leaf image, the more severe the damage from caterpillars; the higher the absorption of the near-infrared band in the leaf image, the more abundant the water content, and the less severe the aphid damage; and the higher the degree of curling in the leaf image, the more severe the damage from parasitic viruses.

4. The method for detecting agricultural pests and diseases according to claim 3, characterized in that: In step S2, when performing multimodal combination of data from visible light and infrared thermal imaging cameras, the specific steps are as follows: S21. Use a bilateral filter to denoise both images while preserving edge information; S22. Perform color correction on the visible light image using adaptive histogram equalization to improve image contrast, and perform brightness matching on the infrared image, using histogram matching to align its brightness distribution with the visible light image. S23. The visible light and infrared images are decomposed into a visual layer containing high-frequency information and a thermal imaging layer containing low-frequency information by using the Laplace pyramid decomposition method, and their respective image features are extracted. S24. In the weighted combination stage, the visual layer and the thermal imaging layer are combined separately. The high-frequency information of the visible light image contains edges and details, showing the outline and defect features of the leaves of legume crops, while the low-frequency information of the infrared image contains the distribution of heat sources, reflecting the infrared spectral characteristics of the leaves of legume crops, which are represented as smooth and significant areas.

5. The method for detecting agricultural pests and diseases according to claim 4, characterized in that: In S24, the combination formula for the visual layer and the thermal imaging layer is: ； in, For position Superimposed brightness at the location, For visible light images at position Brightness at that location For infrared images at location The brightness at each location is calculated, and the two values ​​are then superimposed on the combined image. and For adaptive dynamic weights, the specific weights are determined by the following formula: ， ； In the above formula, The degree of abrupt change in the visible light image. The reflectance is in the near-infrared band, and the degree of abrupt change in the visible light image is determined by position. The rate of change of brightness and color at a location with spatial position A high degree of abrupt change in the visible light image indicates a defect in the leaf at that location. Higher reflectivity in the near-infrared band indicates lower water content in the leaf, thus highlighting the defective and low-water-content areas more prominently.

6. The method for detecting agricultural pests and diseases according to claim 5, characterized in that: In S3, the multimodal combination process is specifically as follows: S31. When determining the confidence level of a point cloud segmentation mask, it is necessary to evaluate the density of points within a cluster. Number of clusters of the same category To determine, if the density of points within a cluster... The higher the value, the more complex the spatial morphology of the clusters, the more pronounced the variation in vertical distance between leaves, the greater the curl of the leaves, and the higher the number of clusters in the same category. The larger the value, the more prevalent this type of cluster is in the current planting area, and therefore the higher the confidence level. For image segmentation masks, the confidence level is determined by the variance of the superimposed pixel values ​​at all locations within the legume leaf image, and the superimposed brightness... The larger the leaf, the more severe the leaf damage, and the higher the confidence level.

7. The method for detecting agricultural pests and diseases according to claim 6, characterized in that: In step S31, the confidence level is calculated as follows: Confidence of point cloud segmentation mask ,in This is the density weighting coefficient. This is the weighting coefficient for the number of clusters. This is a conversion factor for the planting area. The planting area; Confidence of image segmentation mask ,in This represents the number of pixels in an image of a legume leaf. This is the index of a specific pixel. The average superimposed pixel value of all pixels in the image segmentation mask, when Then the point cloud segmentation mask is selected as the final segmentation mask. Then, the image segmentation mask is selected as the final segmentation mask. This is the conversion factor for variance when comparing confidence levels.

8. The method for detecting agricultural pests and diseases according to claim 7, characterized in that: In step S31, in order to reduce the confidence level of the image segmentation mask... The computational complexity during the calculation involves dividing the region of an image segmentation mask into a cross-shaped grid, with the intersection of each grid line serving as the basis for calculating the confidence score. The pixels used in the process are substituted into the formula for calculating the variance. The calculation formula is obtained Approximate value, grid side length of cross grid lines .

Images