Pest control method, device and system based on image recognition and medium

Abstract

The invention belongs to the technical field of pest control, and discloses a pest control method, device and system based on image recognition and a medium. The method comprises the steps of dividing sub-regions, obtaining insect pest type information and insect pest quantity information of each sub-region based on image recognition, analyzing the insect pest type information and the insect pest quantity information of each sub-region, and generating an insect pest prevention and control scheme. The method is applied to pest control, a corresponding pest control scheme is generated through refined early warning of pest types and quantity, automatic deinsectization can be realized, and manual activities are greatly reduced.

Description

Technical Field

[0001] The invention belongs to the technical field of pest control, and discloses a pest control method, equipment, system and medium based on image recognition. Background Art

[0002] Pests and diseases have always been one of the important problems that plague plant cultivation, and have had a certain impact on crop yield and quality. The current pest monitoring methods include manual patrols and universal insecticide devices. Manual patrols are costly, inefficient, and prone to errors, and have a narrow scope of application. For large-scale rotation planting or symbiotic planting, the plants may be very different, and the pests caused are also very different. For example, the size of the pests, flying and crawling, etc., will have a great impact on the choice of insecticide devices.

[0003] Specifically, the following problems will arise: (1) When plants are planted in rotation or at the same time, the types of pests vary. For large plantations, as personnel change, large errors may occur in the method and time of pest control.

[0004] (2) The common image recognition-based pest control systems on the market generally only target insect pests, while ignoring the impact of factors such as water, soil, plant species and climate on pests, leading to incorrect judgments about pests.

[0005] (3) The automation of pest control is low, which either requires a lot of manpower or requires additional monitoring or pest control equipment.

[0006] For example, the patent publication number is CN118154925A, and the disclosed "Pest Assessment and Early Warning Method Based on Multi-Terminal Acquisition and AI Image Recognition Algorithm" includes the following steps: S1, placing multiple cameras at different crop positions in an agricultural factory, monitoring the crops in real time, uploading the monitored video to the image processing model, or taking photos of the crops using a mobile phone APP, and uploading the taken pictures to the image processing model for processing; S2, processing the data uploaded by the camera through edge computing technology, or correcting the photos taken by the mobile phone APP through AI training. The present invention optimizes the algorithm through deep learning technology to improve the accuracy and speed of pest identification, and estimates the level of the current pest by matching the current pest type with the pest type in the database, so that managers can take different treatment methods for different pest levels. Although the invention uses multi-terminal monitoring to warn pests, it is still not combined with agronomy. It is only a simple pest identification. The recognition accuracy is questionable in actual planting scenes. At the same time, the invention uses multi-terminal monitoring, which is costly for scenes such as plantations with multiple planting areas. Summary of the invention

[0007] In view of the deficiencies in the prior art, the present invention provides an insect pest control method, equipment, system and medium based on image recognition.

[0008] In a first aspect, a pest control method based on image recognition comprises: S1, obtaining regional information of plant planting, building a map model based on the regional information; dividing a plurality of sub-regions based on the map model and the regional information of plant planting; the regional information includes the planting area of ​​the plant planting, the map outline of the planting range, the planting distribution, and the monitoring point information of the planting area; S2, acquiring image data in the sub-area, inputting plant information, extracting plant features and pest features from the image data based on the plant information, identifying pest types by combining the pest features with the plant features, and analyzing and predicting the number of pests, so as to obtain pest type information and pest quantity information of the sub-area; S3, repeating step S2 to obtain information on the type and quantity of pests in each sub-area; S4, analyzing the pest type information and pest quantity information of each sub-area and generating a pest control plan.

[0009] Optionally, the step of dividing a plurality of sub-areas based on the map model and the area information of the planted plants in S1 is as follows: The size of the planting area photographed by the monitoring is set based on the ability to identify pest characteristics, and the size of the sub-area is set based on the size of the planting area photographed by the monitoring; Based on the planting distribution, the planting area of ​​the plant is divided into multiple sub-areas with the sub-area size as a unit.

[0010] Optionally, the acquisition of image data in the sub-region in S2, input of plant information, extraction of plant features and pest features from the image data based on the plant information, identification of pest types by combining pest features with plant features, and analysis and prediction of pest quantity to obtain pest type information and pest quantity information of the sub-region, specifically: 21) Selecting a Transformer model, loading the image data in the sub-regions, and inputting plant information into the image data in each sub-region to form mixed image data; the plant information includes the plant species, the planting season, and the climate; 22) Using a Transformer converter in the feature encoding stage of the mixed image data; using a cross sliding window, applying sliding windows in the horizontal and vertical directions on the image, and extracting local features of the mixed image data by adaptively adjusting the stripe width in the horizontal and vertical directions; the local features of the mixed image data include plant features and local features of insect pests; the plant features include plant branch and leaf features; the plant branch and leaf features include the missing state and missing degree of the plant branches and leaves compared to the standardized plant; 23) Reduce the resolution of the feature map layer by layer from the bottom layer to the top layer, aggregate the features at each layer, use the self-attention mechanism to fuse the features at each layer, gradually aggregate the global information, extract the global features at the top layer of the pyramid, and fuse them with the local features of the mixed image data at the bottom layer to obtain the global features of the globally enhanced mixed image data; 24) Mapping the global features of the globally enhanced mixed image data to pest categories and pest quantities using a fully connected layer to obtain pest category information and pest quantity information of the sub-region.

[0011] Optionally, the pest type information and pest quantity information of each sub-region are analyzed in S4 to generate a pest control plan, specifically: 41) Information on pest types includes those that can be eliminated at specific locations and those that have global impact; 42) If the main category of pests is that which can be removed on a spot basis, and the number of pests in a sub-area is sufficient to cause a 10-20% reduction in production, it is determined that the sub-area requires pest treatment; 43) If the main category of pests is global impact, the number of pests in the whole area is sufficient to cause a 1-5% reduction in production, and pest control should be carried out in the whole area.

[0012] In a second aspect, an electronic device includes a processor, wherein the processor is coupled to a memory; The memory stores a computer program that can be loaded by the processor and executes the image recognition-based pest control method described in any one of the first aspects.

[0013] In a third aspect, a computer-readable storage medium includes a computer program or instructions, which, when executed on a computer, enables the computer to execute the pest control method based on image recognition as described in any one of the first aspects.

[0014] In a fourth aspect, an image recognition-based pest control system comprises the electronic device of the second aspect, a slide rail, a plurality of monitoring devices, and a plurality of pest treatment devices.

[0015] Optionally, the slide rail includes an upper slide rail and a lower slide rail; the upper slide rail is installed with a plurality of monitoring devices and surrounds the planting area, so that the monitoring devices can photograph any sub-area; the lower slide rail is installed with a plurality of pest control devices and surrounds the planting area, so that the pest control devices can perform pest control on any sub-area. The present invention has the following advantages: (1) The present invention can automatically analyze the main pests and give pest prompts. If necessary, it can also automatically call the pest removal device, reducing the problems caused by personnel changes or misjudgment. (2) According to different types of plants, targeted image training will be carried out in combination with plant species, parts and planting seasons. At the same time, in the case of plants blocking pests, pests will be analyzed based on the tooth marks left by the plants, greatly improving the accuracy of automatic judgment.

[0016] (3) The slide-type monitoring device can slide at irregular intervals to monitor pests from multiple angles. For large planting areas, monitoring and prevention can be carried out without adding too many devices. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the scheme in the present application, a brief introduction is given below to the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0018] Figure 1 The figure is a flow chart of an insect pest control method based on image recognition described in an embodiment of the present invention.

[0019] Figure 2 The system framework diagram of an insect pest control system based on image recognition described in an embodiment of the present invention.

[0020] Figure 3 This is a globally enhanced layer-by-layer shrinking pyramid structure implementation method for an insect pest control method based on image recognition described in an embodiment of the present invention. DETAILED DESCRIPTION Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by technicians in the technical field to which this application belongs; the terms used in the specification of the application are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" in the specification and claims of this application and the above-mentioned figure descriptions and any variations thereof are intended to cover non-exclusive inclusions.

[0021] In a first aspect, a pest control method based on image recognition comprises: S1, obtaining regional information of plant planting, building a map model based on the regional information; dividing a plurality of sub-regions based on the map model and the regional information of plant planting; the regional information includes the planting area of ​​the plant planting, the map outline of the planting range, the planting distribution, and the monitoring point information of the planting area; S2, acquiring image data in the sub-area, inputting plant information, extracting plant features and pest features from the image data based on the plant information, identifying pest types by combining the pest features with the plant features, and analyzing and predicting the number of pests, so as to obtain pest type information and pest quantity information of the sub-area; S3, repeating step S2 to obtain information on the type and quantity of pests in each sub-area; S4, analyzing the pest type information and pest quantity information of each sub-area and generating a pest control plan.

[0022] Optionally, the step of dividing a plurality of sub-areas based on the map model and the area information of the planted plants in S1 is as follows: The size of the planting area photographed by the monitoring is set based on the ability to identify pest characteristics, and the size of the sub-area is set based on the size of the planting area photographed by the monitoring; Based on the planting distribution, the planting area of ​​the plant is divided into multiple sub-areas with the sub-area size as a unit.

[0023] Through the above scheme, users can more clearly locate the area that needs to be treated for pests, so that when manual pest control activities are required, the pest location can be accurately found; when automated pest control is required, the system can accurately move the pest control device to the area that needs to be treated for pests through the guide rail; at the same time, since the image resolution of different monitoring devices is different, the standard is to be able to identify the characteristics of pests. Specifically, for monitoring devices that use optical magnification, the magnification of the monitoring device is adjusted, and the magnification when the accuracy of identifying the type of pest reaches 90% or above is selected as the standard.

[0024] Optionally, the acquisition of image data in the sub-region in S2, input of plant information, extraction of plant features and pest features from the image data based on the plant information, identification of pest types by combining pest features with plant features, and analysis and prediction of pest quantity to obtain pest type information and pest quantity information of the sub-region, specifically: 21) Selecting a Transformer model, loading the image data in the sub-regions, and inputting plant information into the image data in each sub-region to form mixed image data; the plant information includes the plant species, the planting season, and the climate; 22) Using a Transformer converter in the feature encoding stage of the mixed image data; using a cross sliding window, applying sliding windows in the horizontal and vertical directions on the image, and extracting local features of the mixed image data by adaptively adjusting the stripe width in the horizontal and vertical directions; the local features of the mixed image data include plant features and local features of insect pests; the plant features include plant branch and leaf features; the plant branch and leaf features include the missing state and missing degree of the plant branches and leaves compared to the standardized plant; 23) If Figure 3 As shown in the figure, the resolution of the feature map is reduced layer by layer from the bottom layer to the top layer, the features are aggregated at each layer, and the self-attention mechanism is used to fuse the features at each layer, gradually aggregating the global information. At the top layer of the pyramid, the global features are extracted and fused with the local features of the mixed image data at the bottom layer to obtain the global features of the globally enhanced mixed image data. 24) Mapping the global features of the globally enhanced mixed image data to pest categories and pest quantities using a fully connected layer to obtain pest category information and pest quantity information of the sub-region.

[0025] It should be noted that the computing resources for agricultural image recognition are generally limited, so it is necessary to reduce the amount of calculation and use a hierarchical structure similar to a convolutional network to realize the window division of the image, thereby reducing the computational complexity. First, for an input image of size After the image of patches (each patch is a 4*4 pixel area of ​​the original image) is layered according to the given window size, assuming the window size is M, the amount of calculation can be reduced from the original Reduced to Then, a set of efficient cross sliding windows is constructed to further compress the amount of calculation; after being processed by multiple sets of cross sliding window modules, the generated feature information is input into the feature pyramid network with a global enhancement module, effectively utilizing the advantages of the self-attention mechanism for extracting long-distance context information features; It should be further explained that due to the large relative scale variation of insect pest targets, it is easy to fail to obtain the global features of the target using convolution kernels for feature extraction. The general detection network using feature pyramids can effectively solve this problem.

[0026] Optionally, the pest type information and pest quantity information of each sub-region are analyzed in S4 to generate a pest control plan, specifically: 41) Information on pest types includes those that can be eliminated at specific locations and those that have global impact; 42) If the main category of pests is that which can be removed on a spot basis, and the number of pests in a sub-area is sufficient to cause a 10-20% reduction in production, it is determined that the sub-area requires pest treatment; 43) If the main category of pests is global impact, the number of pests in the whole area is sufficient to cause a 1-5% reduction in production, and pest control should be carried out in the whole area.

[0027] In a second aspect, an electronic device includes a processor, wherein the processor is coupled to a memory; The memory stores a computer program that can be loaded by the processor and executes the image recognition-based pest control method described in any one of the first aspects.

[0028] In a third aspect, a computer-readable storage medium includes a computer program or instructions, which, when executed on a computer, enables the computer to execute the pest control method based on image recognition as described in any one of the first aspects.

[0029] In a fourth aspect, an image recognition-based pest control system includes the electronic device described in the second aspect, a slide rail, a plurality of monitoring devices, and a plurality of pest treatment devices.

[0030] Optionally, the slide rail includes an upper slide rail and a lower slide rail; the upper slide rail is installed with a plurality of monitoring devices and surrounds the planting area, so that the monitoring devices can photograph any sub-area; the lower slide rail is installed with a plurality of pest control devices and surrounds the planting area, so that the pest control devices can perform pest control on any sub-area.

Claims

1. A pest control method based on image recognition, characterized in that: include: S1, obtaining regional information of plant planting, and dividing multiple sub-regions based on the regional information of plant planting; the regional information includes the planting area of ​​the plant planting, the outline of the planting range map, the planting distribution, and the monitoring point information of the planting area; S2, acquiring image data in the sub-area, inputting plant information, extracting plant features and pest features from the image data based on the plant information, identifying pest types by combining the pest features with the plant features, and analyzing and predicting the number of pests, so as to obtain pest type information and pest quantity information of the sub-area; S3, repeating step S2 to obtain information on the type and quantity of pests in each sub-area; S4, analyzing the pest type information and pest quantity information of each sub-area and generating a pest control plan.

2. The method for controlling pests based on image recognition according to claim 1 or 2, characterized in that: In S1, a plurality of sub-areas are divided based on the map model and the regional information of the plant, specifically: The size of the planting area photographed by the monitoring is set based on the ability to identify pest characteristics, and the size of the sub-area is set based on the size of the planting area photographed by the monitoring; Based on the planting distribution, the planting area of ​​the plant is divided into multiple sub-areas with the sub-area size as a unit.

3. The method for pest control based on image recognition according to claim 1, characterized in that: In S2, image data in the sub-region is obtained, plant information is input, plant features and pest features are extracted from the image data based on the plant information, pest types are identified by combining pest features with plant features, and the number of pests is analyzed and predicted to obtain pest type information and pest quantity information in the sub-region, specifically: 21) Selecting a Transformer model, loading the image data in the sub-regions, and inputting plant information into the image data in each sub-region to form mixed image data; the plant information includes the plant species, the planting season, and the climate; 22) Using a Transformer converter in the feature encoding stage of the mixed image data; using a cross sliding window, applying sliding windows in the horizontal and vertical directions on the image, and extracting local features of the mixed image data by adaptively adjusting the stripe width in the horizontal and vertical directions; the local features of the mixed image data include plant features and local features of insect pests; the plant features include plant branch and leaf features; the plant branch and leaf features include the missing state and missing degree of the plant branches and leaves compared to the standardized plant; 23) Reduce the resolution of the feature map layer by layer from the bottom layer to the top layer, aggregate the features at each layer, use the self-attention mechanism to fuse the features at each layer, gradually aggregate the global information, extract the global features at the top layer of the pyramid, and fuse them with the local features of the mixed image data at the bottom layer to obtain the global features of the globally enhanced mixed image data; 24) Mapping the global features of the globally enhanced mixed image data to pest categories and pest quantities using a fully connected layer to obtain pest category information and pest quantity information of the sub-region.

4. The method for pest control based on image recognition according to claim 1, characterized in that: In S4, the pest type information and pest quantity information of each sub-area are analyzed to generate a pest control plan, which is specifically: 41) Information on pest types includes those that can be eliminated at specific locations and those that have global impact; 42) If the main category of pests is that which can be removed on a spot basis, and the number of pests in a sub-area is sufficient to cause a 10-20% reduction in production, it is determined that the sub-area requires pest treatment; 43) If the main category of pests is global impact, the number of pests in the whole area is sufficient to cause a 1-5% reduction in production, and pest control should be carried out in the whole area.

5. An electronic device, characterized in that: A processor is included, wherein the processor is coupled to a memory; The memory stores a computer program that can be loaded by the processor and executes the method according to any one of claims 1 to 4.

6. A computer-readable storage medium, characterized in that: The method comprises a computer program or an instruction, which, when executed on a computer, enables the computer to execute the method according to any one of claims 1 to 4.

7. An insect pest control system based on image recognition, characterized in that: It comprises the electronic device as claimed in claim 6, a slide rail, a plurality of monitoring devices and a plurality of pest control devices.

8. According to the image recognition-based pest control system of claim 8, the slide rail comprises an upper slide rail and a lower slide rail; the upper slide rail is equipped with a plurality of monitoring devices and surrounds the planting area, so that the monitoring devices can photograph any sub-area described in claim 1; the lower slide rail is equipped with a plurality of pest treatment devices and surrounds the planting area, so that the pest treatment devices can perform pest treatment on any sub-area described in claim 1.

Images