Field pest identification and statistics method and system based on image processing

Abstract

The present invention relates to the field of agricultural technologies, and particularly to a method and system for identifying and counting field pests based on image processing. Determine whether each sub-region image is a pest-infected sub-region image or a pest-free sub-region image; analyze the efficacy of the preset type of pesticide applied to each plant according to each pest-infected sub-region image and pest-free sub-region image, so as to obtain the pest control effect of the preset type of pesticide applied to each plant; analyze the characteristics of the pests infected by each plant according to the pest-infected sub-region image of each plant, so as to obtain the pest type of the pests infected by each plant; perform a counting analysis on the pests infected by each plant according to the pest-infected sub-region image of each plant, so as to obtain the pest quantity of the pests infected by each plant and the pest images of each pest. This method realizes the automatic identification and counting of field pests, and significantly improves the efficiency and reliability of the efficacy test.

Description

Technical Field

[0001] The invention relates to the technical field of agriculture, and in particular to a field pest identification and statistical method and system based on image processing. Background Art

[0002] The traditional field efficacy test result survey mainly relies on human observation and inspection records, which requires a lot of manpower and time, especially in large fields. The efficiency is very low and it is difficult to meet the needs of rapid and timely evaluation of efficacy. In addition, the manual identification of pest types and quantities is easily affected by subjective judgment, resulting in inaccurate identification results, making it difficult to accurately count the types and quantities of pests, and affecting the reliability of efficacy evaluation. Due to the lack of accurate statistical data on the types, quantities and distribution of pests, it is difficult to effectively evaluate the effect of pesticide application and determine whether the control effect of pesticides has reached expectations. With the rapid development of agricultural information technology, pest recognition technology based on image processing has gradually become a research hotspot. Image processing technology can be used to automatically identify and count pests, reduce the workload of manual inspections, and improve work efficiency. However, most of the existing pest recognition methods based on image processing focus on the identification of specific pests, lack the identification and statistical analysis of multiple pests, and are difficult to effectively evaluate the effect of pesticide application. In view of this, the present application proposes a field pest recognition and statistical method and system based on image processing. Summary of the invention

[0003] The present invention overcomes the deficiencies of the prior art and provides a field pest identification and statistical method and system based on image processing.

[0004] To achieve the above-mentioned purpose, the technical solution adopted by the present invention is:

[0005] The first aspect of the present invention discloses a field pest identification and statistical method based on image processing, comprising the following steps:

[0006] Obtain characteristic images of various types of pests in the field, and construct a pest characteristic knowledge graph based on the characteristic images of various types of pests;

[0007] Based on the plant images of each plant in the field after the pesticide is applied, the plant images are obtained by the camera device, the plant images are divided into a plurality of sub-region images, and each sub-region image is judged to be an image of a sub-region infected by insect pests or an image of a sub-region not infected by insect pests;

[0008] Analyze the properties of the preset type of pesticide applied to each plant based on the images of each pest-infected sub-region and the images of the non-pest-infected sub-region to obtain the pest control effect of the preset type of pesticide applied to each plant;

[0009] Performing feature analysis on the pests infected by each plant according to the pest-infected sub-region images of each plant, and obtaining the pest types of the pests infected by each plant;

[0010] Count and analyze the pests infested on each plant based on the pest-infected sub-region images of each plant, to obtain the pest quantity of the pests infested on each plant and the pest images of each pest.

[0011] Generate a statistical analysis report for each plant based on the pest control effect of the preset type of pesticide applied to each plant, the pest type of the pests infested, the pest quantity of the pests infested, and the pest images of each pest, and send the statistical analysis report to a preset terminal.

[0012] Preferably, obtain the characteristic images of various types of pests in the field, and construct a pest characteristic knowledge graph according to the characteristic images of various types of pests, specifically:

[0013] Obtain all possible various types of pests in the field, and obtain the characteristic images of various types of pests through a big data network;

[0014] Construct a knowledge graph, and divide the knowledge graph into several graph sub-nodes based on all possible various types of pests in the field, and label an index label of a type of pest for each graph sub-node according to various types of pests;

[0015] Store the characteristic images of various types of pests on the graph sub-nodes labeled with the corresponding index labels of the types of pests respectively, to obtain a pest characteristic knowledge graph.

[0016] Preferably, based on the camera device, obtain the plant images of each plant in the field after applying pesticides, divide the plant images into several sub-region images, and determine whether each sub-region image is a pest-infected sub-region image or a pest-free sub-region image, specifically:

[0017] After applying the preset type of pesticide to the plants in the field respectively, based on the plant images of the plants in the field after applying pesticides obtained by the camera device, introduce a voxelization algorithm, and preset the size of the voxel grid, and voxelize the plant images into several voxel grids based on the voxelization algorithm;

[0018] Obtain the regional images corresponding to each voxel grid to obtain several sub-region images; randomly access any sub-region image, and calculate the mean square error value between the sub-region image and the characteristic image stored on the graph sub-node in the pest characteristic knowledge graph;

[0019] Determine the similarity between the sub-region image and the characteristic images on each graph sub-node according to the mean square error value between the sub-region image and the characteristic image stored on the graph sub-node in the pest characteristic knowledge graph;

[0020] The similarities between the sub-region image and the feature images on each atlas sub-node are compared with a preset similarity threshold. If the similarities between the sub-region image and the feature images on each atlas sub-node are not greater than the preset similarity threshold, it means that there is no insect pest in the sub-region image, and the sub-region image is marked as an insect pest-free sub-region image;

[0021] If the similarity between the sub-region image and a feature image on a certain atlas sub-node is greater than a preset similarity threshold, it indicates that there is pest in the sub-region image, and the sub-region image is marked as a pest-infected sub-region image;

[0022] Visit the next unvisited sub-region image, repeat the above steps to determine whether the sub-region image is an insect pest infected sub-region image or an insect pest free sub-region image, and so on, until all sub-region images are determined and analyzed.

[0023] Preferably, the properties of the preset type of pesticide applied to each plant are analyzed according to each pest-infected sub-region image and the non-pest-infected sub-region image to obtain the pest control effect of the preset type of pesticide applied to each plant, specifically:

[0024] Calculating the total number of all images marked as pest-infected sub-regions and images marked as non-pest-infected sub-regions, and calculating the percentage of the pest-infected area of ​​the plant according to the total number of all images marked as pest-infected sub-regions and images marked as non-pest-infected sub-regions;

[0025] Comparing the percentage value of the insect pest infection area of ​​the plant with a preset percentage value;

[0026] If the percentage value of the insect pest infection area of ​​the plant is not greater than the preset percentage value, the preset type of pesticide applied to the plant is obtained, indicating that the pest control effect of the preset type of pesticide meets the standard, and the preset type of pesticide applied to the plant is calibrated as a type of preset type of pesticide;

[0027] If the proportion of the insect pest infection area of ​​the plant is greater than the preset proportion, the preset type of pesticide applied to the plant is obtained, indicating that the pest control effect of the preset type of pesticide does not meet the standard, and the preset type of pesticide applied to the plant is calibrated as a type II preset type pesticide.

[0028] Preferably, the pests infected by each plant are analyzed for their characteristics according to the pest-infected sub-region images of each plant, so as to obtain the pest types of the pests infected by each plant, specifically:

[0029] If the similarity between the sub-region image and a feature image on a certain atlas sub-node is greater than a preset similarity threshold, then the index label marked on the atlas sub-node corresponding to the similarity greater than the preset similarity threshold is obtained;

[0030] Determine the type of pest existing in the corresponding pest-infected sub-region image based on the index label calibrated for the graph sub-node corresponding to the similarity greater than the preset similarity threshold;

[0031] And so on, integrate and process the types of pests existing in all pest-infected sub-region images of each plant to obtain the types of pests infected in each plant.

[0032] Preferably, perform a count analysis on the pests infected in each plant according to the pest-infected sub-region images of each plant to obtain the number of pests infected in each plant and the pest images of each individual pest, specifically as follows:

[0033] Obtain the pest-infected sub-region images of each plant, and perform grayscale conversion, Gaussian filtering, and contrast enhancement processing on the pest-infected sub-region images;

[0034] Use the Prewitt operator algorithm to calculate the gradient value of each pixel point in the pest-infected sub-region image, and calibrate the pixel point with the maximum gradient value as the seed point;

[0035] Preset the size of the domain range, and calculate the difference in grayscale values between the seed point and the neighboring pixel points within its preset domain range;

[0036] If the difference in grayscale values between a certain neighboring pixel point and the seed point is greater than the preset difference threshold, then mark the neighboring pixel point as an affiliated pixel point and add the affiliated pixel point to the domain to which the seed point belongs;

[0037] If the difference in grayscale values between a certain neighboring pixel point and the seed point is not greater than the preset difference threshold, then use the neighboring pixel point as a new seed point;

[0038] Repeat the above steps to add each pixel point to the domain to which the seed point or the new seed point belongs, and continuously iterate until no new pixel points can be added to the domain to which the seed point or the new seed point belongs, to obtain several seed regions;

[0039] Obtain the image information of each seed region to separate a single pest in the pest-infected sub-region image from the image background, and obtain the pest images of each individual pest in each plant;

[0040] Count the regional quantity value of the seed regions, and use the regional quantity value of the seed regions as the number of pests infected in the corresponding plant.

[0041] The imaging device includes:

[0042] A camera, responsible for collecting images of field plants;

[0043] The adjustment and support module is responsible for fixing the camera and adjusting the shooting angle of the camera;

[0044] The data storage module stores image data;

[0045] The power supply module provides power supply for the whole system.

[0046] In the second aspect of the present invention, a field pest identification and statistics system based on image processing is disclosed. The field pest identification and statistics system based on image processing includes a memory and a processor. A program of the field pest identification and statistics method based on image processing is stored in the memory. When the program of the field pest identification and statistics method based on image processing is executed by the processor, the steps of any one of the field pest identification and statistics methods based on image processing are realized.

[0047] In the third aspect of the present invention, a computer-readable storage medium is disclosed. The computer-readable storage medium includes a program of the field pest identification and statistics method based on image processing. When the program of the field pest identification and statistics method based on image processing is executed by a processor, the steps of any one of the field pest identification and statistics methods based on image processing are realized.

[0048] The present invention solves the technical defects existing in the background technology. The present invention has the following beneficial effects: based on the plant images of each plant in the field after pesticide application obtained by the imaging device, the plant images are segmented into several sub-region images, and it is judged whether each sub-region image is a pest-infected sub-region image or a pest-free sub-region image; according to each pest-infected sub-region image and pest-free sub-region image, the drug properties of the preset type of pesticide applied to each plant are analyzed to obtain the pest control effect of the preset type of pesticide applied to each plant; according to the pest-infected sub-region images of each plant, the characteristics of the pests infected by each plant are analyzed to obtain the pest types of the pests infected by each plant; according to the pest-infected sub-region images of each plant, the counting analysis of the pests infected by each plant is carried out to obtain the pest quantity of the pests infected by each plant and the pest images of each pest; according to the pest control effect of the preset type of pesticide applied to each plant, the pest types of the pests infected by each plant, the pest quantity of the pests infected by each plant, and the pest images of each pest, a statistical analysis report of each plant is generated. This method realizes the automatic identification and statistics of field pests, and significantly improves the efficiency and reliability of the pesticide efficacy test. Description of the Drawings

[0049] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained according to these drawings.

[0050] Figure 1 This is a flow chart of the first method of the statistical method for pest identification in Honda;

[0051] Figure 2 The flowchart of the second method of the statistical method for pest identification in Honda;

[0052] Figure 3 It is a simplified structural diagram of the camera equipment. DETAILED DESCRIPTION

[0053] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0054] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0055] The first aspect of the present invention discloses a statistical method for identifying field pests based on image processing, such as Figure 1 As shown, the following steps are included:

[0056] S102: Acquire characteristic images of various types of insect pests in the field, and construct an insect pest characteristic knowledge graph based on the characteristic images of various types of insect pests;

[0057] S104: based on the plant images of each plant in the field after the pesticide application obtained by the camera device, the plant images are divided into a plurality of sub-region images, and each sub-region image is judged to be an image of a sub-region infected with insect pests or an image of a sub-region not infected with insect pests;

[0058] S106: analyzing the properties of the preset type of pesticide applied to each plant according to the images of each pest-infected sub-region and the images of the non-pest-infected sub-region, to obtain the pest control effect of the preset type of pesticide applied to each plant;

[0059] S108: performing feature analysis on the pests infected by each plant according to the pest-infected sub-region image of each plant, and obtaining the pest type of the pests infected by each plant;

[0060] S110: Counting and analyzing the pests infected by each plant according to the pest infected sub-region image of each plant, and obtaining the number of pests infected by each plant and the pest image of each pest;

[0061] S112: Generate a statistical analysis report for each plant based on the pest control effect of the preset type of pesticide applied to each plant, the type of pest infestation, the number of pests infested, and the pest images of each pest, and send the statistical analysis report to a preset terminal.

[0062] It should be noted that, first of all, by obtaining the characteristic images of various types of pests and constructing a pest characteristic knowledge graph, a visual recognition basis for different pests is established. Then, the plant images in the field are obtained by using a camera device after pesticide application, and these images are segmented. The plant images are decomposed into multiple sub-region images. The system will further determine whether each sub-region is infested by pests, which is carried out by comparing the sub-region images with the known pest characteristic knowledge graph. This step helps to accurately locate the occurrence position of pests and improve the recognition accuracy. Then, by analyzing the pest-infested sub-regions and the non-pest-infested sub-regions, the efficacy of the preset type of pesticide can be evaluated. This includes not only the direct effect of the drug properties, but also may involve comprehensive evaluation indicators such as the stability, permeability, and insecticidal efficiency of the pesticide in a specific environment, so as to provide a basis for the selection and use of pesticides. Next, the system conducts a characteristic analysis of the pests to determine the specific type of pest infested in each plant. Subsequently, a count analysis of the number of pests is carried out to quantify the distribution of pests. Combining the acquisition of pest images, more intuitive pest examples can be obtained, which is convenient for further analysis and research. Finally, the above analysis results are summarized to generate a statistical analysis report and pushed through a preset terminal (such as a mobile phone, a computer, etc.). This not only facilitates agricultural managers to timely understand the pest situation in the field, but also provides detailed data support for subsequent decision-making, such as adjusting the pesticide use strategy and optimizing the planting plan. In summary, this method realizes the automatic recognition and statistics of field pests by integrating advanced technologies such as image processing and machine learning, significantly improves the efficiency and reliability of the efficacy test, and provides a scientific, efficient, and accurate decision-making support tool for agricultural production.

[0063] Preferably, obtain the characteristic images of various types of pests in the field, and construct a pest characteristic knowledge graph according to the characteristic images of various types of pests, as Figure 2 shown, specifically:

[0064] S202: Obtain all possible various types of pests in the field, and obtain the characteristic images of various types of pests through a big data network;

[0065] S204: Construct a knowledge graph, and based on all possible various types of pests in the field, divide the knowledge graph into several graph sub-nodes, and label each graph sub-node with an index label of a type of pest according to various types of pests;

[0066] S206: Store the characteristic images of various types of insect pests on the atlas sub-nodes with the index labels calibrated for the corresponding types of insect pests respectively, to obtain the insect pest characteristic knowledge atlas.

[0067] It should be noted that all possible samples of various types of insect pests are obtained, and the characteristic images of these insect pests are obtained by using the big data network. These characteristic images contain various appearance characteristics of the insect pests, such as size, shape, color, texture, etc., laying a foundation for subsequent identification and classification. Based on the large number of collected insect pest samples, a comprehensive knowledge atlas is constructed. This atlas divides all possible types of insect pests into several atlas sub-nodes, and each sub-node represents a specific type of insect pest. This structured way enables the atlas to have a global perspective and at the same time be able to focus on specific insect pest species. For the convenience of management and query, each atlas sub-node is assigned an index label of the type of insect pest. In this way, during the actual operation process, when an insect pest sample is obtained, it can be quickly located to the corresponding atlas sub-node through its appearance characteristics, and then the type of the insect pest can be identified. Store the characteristic images of each type of insect pest on the atlas sub-node corresponding to its index label. This precise image-label correspondence makes the knowledge atlas a powerful database, which can not only be used to identify unknown insect pests, but also provide rich reference information, such as the life cycle of the insect pest, control methods, etc. The insect pest characteristic knowledge atlas constructed in this way can greatly improve the accuracy and efficiency of insect pest identification.

[0068] Preferably, based on the plant images of each plant in the field after pesticide application obtained by the imaging device, divide the plant images into several sub-region images, and judge whether each sub-region image is an insect pest-infected sub-region image or a non-insect pest-infected sub-region image. Specifically:

[0069] After applying a preset type of pesticide to the plants in the field respectively, based on the plant images of the plants in the field after pesticide application obtained by the imaging device, introduce a voxelization algorithm, and preset the size of the voxel grid. Based on the voxelization algorithm, voxelize the plant images into several voxel grids;

[0070] Obtain the regional images corresponding to each voxel grid to obtain several sub-region images; randomly access any sub-region image, and calculate the mean square error value between the sub-region image and the characteristic images stored on the atlas sub-nodes in the insect pest characteristic knowledge atlas;

[0071] Determine the similarity between the sub-region image and the characteristic images on each atlas sub-node according to the mean square error value between the sub-region image and the characteristic images stored on the atlas sub-nodes in the insect pest characteristic knowledge atlas;

[0072] The similarities between the sub-region image and the feature images on each atlas sub-node are compared with a preset similarity threshold. If the similarities between the sub-region image and the feature images on each atlas sub-node are not greater than the preset similarity threshold, it means that there is no insect pest in the sub-region image, and the sub-region image is marked as an insect pest-free sub-region image;

[0073] If the similarity between the sub-region image and a feature image on a certain atlas sub-node is greater than a preset similarity threshold, it indicates that there is pest in the sub-region image, and the sub-region image is marked as a pest-infected sub-region image;

[0074] Visit the next unvisited sub-region image, repeat the above steps to determine whether the sub-region image is an insect pest infected sub-region image or an insect pest free sub-region image, and so on, until all sub-region images are determined and analyzed.

[0075] It should be noted that after applying the preset type of pesticide to the plants in the field, the plant images are collected using a camera device, and the voxelization algorithm is introduced to divide the plant images into several voxel grids. The size of the voxel grid is a pre-set parameter. This step helps to refine the image processing and improve the recognition accuracy. The corresponding regional image is extracted from each voxel grid to form a series of sub-region images, which is the basis for further pest identification. For each sub-region image, the mean square error value between it and the feature image stored in the pest feature knowledge graph is calculated. The smaller the mean square error value, the higher the similarity between the two images. Based on the calculated similarity value, it is judged whether the sub-region image matches the pest feature. If the similarity is lower than the preset threshold, it means that there is no pest in the area, and it is marked as a pest-free sub-region image; otherwise, it is marked as a pest-infected sub-region image. Repeat the above steps for all sub-region images that have not been visited until all sub-region images are identified. It can accurately identify the distribution location of pests on plants, provide real-time and accurate pest information, and provide strong support for the automated monitoring and management of farmland pests.

[0076] Preferably, the properties of the preset type of pesticide applied to each plant are analyzed according to each pest-infected sub-region image and the non-pest-infected sub-region image to obtain the pest control effect of the preset type of pesticide applied to each plant, specifically:

[0077] Calculating the total number of all images marked as pest-infected sub-regions and images marked as non-pest-infected sub-regions, and calculating the percentage of the pest-infected area of ​​the plant according to the total number of all images marked as pest-infected sub-regions and images marked as non-pest-infected sub-regions;

[0078] Comparing the percentage value of the insect pest infection area of ​​the plant with a preset percentage value;

[0079] If the ratio of the pest-infected area of the plant is not greater than the preset ratio, obtain the preset type of pesticide applied to the plant, indicate that the pest control effect of the preset type of pesticide meets the standard, and label the preset type of pesticide applied to the plant as a type-I preset type of pesticide;

[0080] If the ratio of the pest-infected area of the plant is greater than the preset ratio, obtain the preset type of pesticide applied to the plant, indicate that the pest control effect of the preset type of pesticide does not meet the standard, and label the preset type of pesticide applied to the plant as a type-II preset type of pesticide.

[0081] It should be noted that first, the plant image is segmented into a pest-infected sub-region image and a non-pest-infected sub-region image through image processing technology. Next, the ratio of the number of pest-infected sub-region images to the total number of regions is statistically calculated to obtain the ratio of the pest-infected area. This ratio reflects the degree of pest coverage on the plant after pesticide application. The calculated ratio of the pest-infected area is compared with a preset reference value. The preset reference value usually represents the ideal pest control effect standard. If the actual ratio is less than or equal to the preset reference value, it indicates that the pest control effect of the pesticide meets the standard, and it can be classified as a "type-I preset type of pesticide". If the actual ratio is greater than the preset reference value, the pest control effect of the pesticide does not meet the expectation, and it can be classified as a "type-II preset type of pesticide".

[0082] Preferably, the pests infected by each plant are analyzed for characteristics based on the pest-infected sub-region images of each plant to obtain the pest types of the pests infected in each plant, specifically:

[0083] If the similarity between the sub-region image and the characteristic image on a certain atlas sub-node is greater than the preset similarity threshold, obtain the index label calibrated for the atlas sub-node corresponding to the similarity greater than the preset similarity threshold;

[0084] Determine the pest type existing in the corresponding pest-infected sub-region image according to the index label calibrated for the atlas sub-node corresponding to the similarity greater than the preset similarity threshold;

[0085] And so on, the pest types existing in all pest-infected sub-region images in each plant are integrally processed to obtain the pest types of the pests infected in each plant.

[0086] It should be noted that, first, images of field plants after pesticide application are obtained by a camera device, and the images are segmented into multiple sub-region images using image processing technology. This step is based on the assumption that pests are mainly concentrated in specific parts of the plants, so segmentation can more accurately locate and identify pests. For each sub-region image, the system calculates the similarity between it and the feature images in the pest feature knowledge graph. The pest feature knowledge graph contains feature images of various known pests. By comparing the similarity between the sub-region image and these feature images, it can be preliminarily determined whether there is a certain type of pest in the sub-region image. A preset similarity threshold is set. If the similarity between the sub-region image and the feature image on a certain graph sub-node exceeds this threshold, it indicates that there is likely a pest type corresponding to the feature image in the sub-region image. Once it is determined that the similarity between the sub-region image and the feature image of a certain graph sub-node is high, the system determines the pest type existing in the sub-region image according to the indexed label calibrated for the graph sub-node. The above analysis is performed on the pest-infected sub-region images of all plants, and after integration processing, the pest types infected in each plant are obtained. This method can automatically identify the pest types on field plants, greatly improving the efficiency and accuracy of identification and reducing the error rate of manual observation and identification.

[0087] Preferably, count analysis is performed on the pests infected in each plant according to the pest-infected sub-region images of each plant to obtain the pest quantity of the pests infected in each plant and the pest images of each pest, specifically:

[0088] Obtain the pest-infected sub-region images of each plant, and perform grayscale conversion, Gaussian filtering, and contrast enhancement processing on the pest-infected sub-region images;

[0089] Use the Prewitt operator algorithm to calculate the gradient value of each pixel point in the pest-infected sub-region image, and calibrate the pixel point with the largest gradient value as the seed point;

[0090] Preset the size of the domain range, and calculate the difference in grayscale values between the seed point and the neighboring pixel points within its preset domain range;

[0091] If the difference in grayscale values between a certain neighboring pixel point and the seed point is greater than the preset difference threshold, then mark the neighboring pixel point as an affiliated pixel point and add the affiliated pixel point to the domain to which the seed point belongs;

[0092] If the difference in grayscale values between a certain neighboring pixel point and the seed point is not greater than the preset difference threshold, then use the neighboring pixel point as a new seed point;

[0093] Repeat the above steps to add each pixel point to the neighborhood of the seed point or the new seed point, and continuously iterate until no new pixel points can be added to the neighborhood of the seed point or the new seed point, obtaining several seed regions;

[0094] Obtain the image information of each seed region to separate single pests in the pest-infected sub-region image from the image background, and obtain the pest images of each pest in each plant;

[0095] Count the regional quantity value of the seed region, and use the regional quantity value of the seed region as the pest quantity of the pests infected in the corresponding plant.

[0096] It should be noted that, first, the obtained pest-infected sub-region image is subjected to grayscale processing, Gaussian filtering, and contrast enhancement processing. Grayscale processing simplifies the color information of the image, Gaussian filtering removes noise, and contrast enhancement highlights the details of the image, laying a foundation for subsequent image analysis. The Prewitt operator algorithm is used to calculate the gradient value of each pixel point in the image, and the pixel point with the largest gradient value is found as the seed point. The seed point is usually a key feature point on the edge of the pest. A preset neighborhood range size is set, and the grayscale value difference between the seed point and its neighboring pixel points is calculated. If the difference exceeds the preset threshold, the neighboring pixel point is marked as an affiliated pixel point and added to the neighborhood of the seed point. This process is continuously iterated until no new pixel points can be added to the neighborhood. Through the above steps, multiple seed regions are finally obtained, and each region corresponds to a single pest in the image. These regions clearly isolate the pests, facilitating subsequent analysis and statistics. The image information of a single pest is extracted from the seed region, that is, the pest images of each pest in each plant. The total number of seed regions is counted, that is, the number of pests. Each seed region represents a single pest, so the number of seed regions directly corresponds to the number of pests. Through seed point detection and neighborhood range expansion, this method can accurately identify single pests in the image and can effectively distinguish them even in the case of dense distribution. The entire process has a high degree of automation, reduces manual intervention, and improves efficiency and consistency. It can accurately count the number of pests on each plant and can also obtain the image information of the pests, providing basic data for subsequent pest analysis and research.

[0097] In addition, this method further includes the following steps:

[0098] Obtain the chemical composition information of various preset types of pesticides, and generate a retrieval label according to the chemical composition information of various preset types of pesticides;

[0099] Based on the retrieval label, retrieve the big data network to obtain the volatilization rate of various preset types of pesticides under various combinations of environmental factors;

[0100] Construct a knowledge base, and import the volatilization rates of various pesticides of preset types under various combinations of environmental factors into the knowledge base to obtain a volatilization rate knowledge base;

[0101] Before applying pesticides of a preset type to plants in the field, obtain the latest pesticide application time node for each plant to apply pesticides of the preset type;

[0102] Based on the latest pesticide application time node for each plant to apply pesticides of the preset type, formulate a preset pesticide application time period for each plant to apply pesticides of the preset type;

[0103] Based on weather prediction software, obtain the predicted environmental factors for applying pesticides of a preset type to each plant during each preset pesticide application time period;

[0104] Import the predicted environmental factors for applying pesticides of a preset type to each plant during each preset pesticide application time period into the volatilization rate knowledge base for pairing, and obtain the predicted volatilization rates after applying pesticides of a preset type to each plant during each preset pesticide application time period;

[0105] Sort the predicted volatilization rates after applying pesticides of a preset type to each plant during each preset pesticide application time period, extract the preset pesticide application time period corresponding to the minimum predicted volatilization rate, and output the preset pesticide application time period corresponding to the minimum predicted volatilization rate as the optimal pesticide application time period for applying pesticides of the preset type to the corresponding plant;

[0106] By analogy, obtain the optimal pesticide application time period for applying pesticides of a preset type to each plant.

[0107] It should be noted that, first of all, the system obtains and analyzes the information of various preset types of pesticide ingredients to generate retrieval tags for searching and classification. Using the big data network and the retrieval tags, the system searches for and obtains the volatilization rate data of various preset types of pesticides under different environmental conditions. These data include factors such as temperature, humidity, and wind speed. The collected volatilization rate data is integrated into the knowledge base to form a database on the relationship between pesticide volatilization rate and environmental factors, facilitating subsequent queries and applications. Before pesticide application, the system determines the latest pesticide application time point for each plant according to the specific needs of each plant (for example, considering the crop growth cycle, peak period of pests and diseases, etc.). This is to ensure pesticide application during the critical period of crop growth, while avoiding being too early or too late to improve the pesticide efficacy. Based on the weather prediction software, the system obtains the predicted environmental factors such as temperature, humidity, and wind speed during the predicted pesticide application period to provide real-time or future environmental conditions for volatilization rate prediction. The predicted environmental factors are imported into the volatilization rate knowledge base, and the system calculates the volatilization rate after pesticide application during the predicted pesticide application period according to historical data and current environmental prediction. This step uses machine learning or complex algorithms to simulate the behavior of pesticides in a specific environment. The system sorts the predicted volatilization rates and finds the predicted pesticide application period corresponding to the minimum volatilization rate as the optimal pesticide application time. This ensures that pesticides exert the maximum efficacy in the most favorable environment while reducing the negative impact on the environment. Through precise calculation and prediction, the optimal pesticide application time is selected for each experimental plant to maximize the efficacy of pesticides and minimize environmental impact, thereby improving the reliability of experimental results and enabling the provision of more reliable evaluation reports.

[0108] In addition, in the steps after obtaining the plant images of each plant in the field after pesticide application based on the imaging device, the following steps may further be included:

[0109] Introduce a locally linear embedding model, input the plant image into the locally linear embedding model, perform feature decomposition to obtain an orthogonal matrix and a diagonal matrix composed of linear vectors,

[0110] Multiply the orthogonal matrix and the diagonal matrix to obtain a feature vector matrix;

[0111] Select the eigenvector with the largest eigenvalue in the feature vector matrix as the reference point, and construct a new coordinate system based on the reference point;

[0112] Map all the eigenvectors in the feature vector matrix into the new coordinate system to generate a target point cloud data matrix; the target point cloud data matrix contains the position information of all the eigenvectors in the image in the new coordinate system.

[0113] Extract the coordinate points of each extreme point from the target point cloud data matrix to obtain a set of extreme coordinate points, and the extreme coordinate points represent the most important feature information in the image;

[0114] Recombine the set of extreme coordinate points into the world coordinate system to generate a denoised plant image.

[0115] It should be noted that plant images in the field after spraying are captured by a camera device. These images contain the appearance features of the plants, including possible pest infestations, etc. Next, a locally linear embedding model is introduced. This model decomposes the image features into an orthogonal matrix composed of linear vectors and a diagonal matrix by analyzing the local linear relationships in the image. The orthogonal matrix represents the orthogonal basis of the eigenvectors, while the diagonal matrix records the weights or importance of these eigenvectors. By multiplying the orthogonal matrix and the diagonal matrix, an eigenvector matrix is obtained, which contains all the eigenvectors and their weights and is used for subsequent image analysis and processing. Select the eigenvector with the largest eigenvalue from the eigenvector matrix as the reference point, and based on this reference point, a new coordinate system is constructed, which can more effectively represent the key information in the image. Map all the eigenvectors into the new coordinate system to generate a target point cloud data matrix, which contains the position information of all the eigenvectors in the new coordinate system and helps to more clearly display the key features in the image. Extract the coordinates of each extreme point from the target point cloud data matrix to form a set of extreme coordinate points. These extreme points represent the most important or significant feature information in the image. Recombine the set of extreme coordinate points into the world coordinate system to generate a denoised plant image. This process helps to highlight the key features of the plants while reducing the influence of noise and other irrelevant information. Through the locally linear embedding model, the system can more accurately identify and extract the key features in the image, such as the shape and size of the plants and possible pest infestations. The generated denoised plant image helps to remove noise and other interference factors in the image, making the key information more clearly visible.

[0116] As Figure 3 shown, the camera device includes:

[0117] A camera 2011, responsible for collecting plant images in the field;

[0118] An adjustment support module 2022, responsible for fixing the camera and adjusting the shooting angle of the camera;

[0119] A data storage module 2033, storing image data;

[0120] A power supply module 2044, providing power supply for the entire system.

[0121] In a second aspect of the present invention, a field pest identification and statistics system based on image processing is disclosed. The field pest identification and statistics system based on image processing includes a memory and a processor. A program for the field pest identification and statistics method based on image processing is stored in the memory. When the program for the field pest identification and statistics method based on image processing is executed by the processor, the steps of any of the field pest identification and statistics methods based on image processing are implemented.

[0122] In a third aspect of the present invention, a computer-readable storage medium is disclosed. The computer-readable storage medium includes a program for the field pest identification and statistics method based on image processing. When the program for the field pest identification and statistics method based on image processing is executed by a processor, the steps of any of the field pest identification and statistics methods based on image processing are implemented.

[0123] In several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined, or can be integrated into another system, or some features can be ignored, or not executed. Additionally, the couplings, or direct couplings, or communication connections between the various components shown or discussed can be through some interfaces. The indirect couplings or communication connections of devices or units can be electrical, mechanical, or other forms.

[0124] The units described above as separate components may or may not be physically separated. The components shown as units may or may not be physical units. They can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0125] Furthermore, in each embodiment of the present invention, the various functional units can all be integrated in one processing unit, or each unit can be separately regarded as one unit, or two or more units can be integrated in one unit. The above-mentioned integrated units can be implemented in the form of hardware, or in the form of a combination of hardware and software functional units.

[0126] Those of ordinary skill in the art can understand that all or part of the steps to implement the above method embodiments can be completed by hardware related to program instructions. The foregoing program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps including the above method embodiments; and the foregoing storage medium includes: various media that can store program codes such as removable storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs.

[0127] Alternatively, if the above integrated unit of the present invention is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the embodiments of the present invention, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the methods of the various embodiments of the present invention. And the foregoing storage medium includes: various media that can store program codes such as removable storage devices, ROM, RAM, magnetic disks, or optical discs.

[0128] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention.

Claims

1. A statistical method for identifying field pests based on image processing, characterized in that: The following steps are involved: Obtain characteristic images of various types of pests in the field, and construct a pest characteristic knowledge graph based on the characteristic images of various types of pests; Based on the plant images of each plant in the field after the pesticide is applied, the plant images are obtained by the camera device, the plant images are divided into a plurality of sub-region images, and each sub-region image is judged to be an image of a sub-region infected by insect pests or an image of a sub-region not infected by insect pests; Analyze the properties of the preset type of pesticide applied to each plant based on the images of each pest-infected sub-region and the images of the non-pest-infected sub-region to obtain the pest control effect of the preset type of pesticide applied to each plant; Performing feature analysis on the pests infected by each plant according to the pest-infected sub-region images of each plant, and obtaining the pest types of the pests infected by each plant; Count and analyze the pests infected by each plant according to the pest infection sub-region image of each plant, and obtain the number of pests infected by each plant and the pest image of each pest; A statistical analysis report for each plant is generated based on the pest control effect of the preset type of pesticide applied to each plant, the pest type of the infected pests, the number of infected pests and the pest image of each pest, and the statistical analysis report is sent to a preset terminal.

2. The method for identifying field pests based on image processing according to claim 1, characterized in that: Obtain the characteristic images of various types of pests in the field, and build a pest characteristic knowledge graph based on the characteristic images of various types of pests, specifically: Obtain all possible types of pests in the field and obtain characteristic images of various types of pests through the big data network; Constructing a knowledge graph, and dividing the knowledge graph into a plurality of graph sub-nodes based on all possible types of pests in the field, and labeling each graph sub-node with an index label of a type of pest according to the type of pest; The characteristic images of various types of pests are stored in the graph sub-nodes marked with index labels of the corresponding types of pests, and the pest characteristic knowledge graph is obtained.

3. The method for identifying field pests based on image processing according to claim 1, characterized in that: Based on the plant images of each plant in the field after the application of pesticides obtained by the camera equipment, the plant images are divided into a plurality of sub-region images, and it is determined whether each sub-region image is an image of a pest-infected sub-region or an image of a non-pest-infected sub-region, specifically: After applying preset types of pesticides to plants in the field, plant images of the plants in the field after the pesticide application are obtained based on a camera device, a voxelization algorithm is introduced, and the size of the voxel grid is preset, and the plant image is voxelized into a plurality of voxel grids based on the voxelization algorithm; Obtaining the region image corresponding to each voxel grid to obtain a plurality of sub-region images; randomly accessing any sub-region image, and calculating the mean square error value between the sub-region image and the feature image stored on the graph sub-node in the pest feature knowledge graph; Determine the similarity between the sub-region image and the feature images on each graph sub-node according to the mean square error value between the sub-region image and the feature images stored on the graph sub-node in the pest feature knowledge graph; The similarities between the sub-region image and the feature images on each atlas sub-node are compared with a preset similarity threshold. If the similarities between the sub-region image and the feature images on each atlas sub-node are not greater than the preset similarity threshold, it means that there is no insect pest in the sub-region image, and the sub-region image is marked as an insect pest-free sub-region image; If the similarity between the sub-region image and a feature image on a certain atlas sub-node is greater than a preset similarity threshold, it indicates that there is pest in the sub-region image, and the sub-region image is marked as a pest-infected sub-region image; Visit the next unvisited sub-region image, repeat the above steps to determine whether the sub-region image is an insect pest infected sub-region image or an insect pest free sub-region image, and so on, until all sub-region images are determined and analyzed.

4. The method for identifying field pests based on image processing according to claim 1, characterized in that: The properties of the preset types of pesticides applied to each plant are analyzed based on the images of each pest-infected sub-region and the images of the non-pest-infected sub-region, and the pest control effects of the preset types of pesticides applied to each plant are obtained, specifically: Calculating the total number of all images marked as pest-infected sub-regions and images marked as non-pest-infected sub-regions, and calculating the percentage of the pest-infected area of ​​the plant according to the total number of all images marked as pest-infected sub-regions and images marked as non-pest-infected sub-regions; Comparing the percentage value of the insect pest infection area of ​​the plant with a preset percentage value; If the percentage value of the insect pest infection area of ​​the plant is not greater than the preset percentage value, the preset type of pesticide applied to the plant is obtained, indicating that the pest control effect of the preset type of pesticide meets the standard, and the preset type of pesticide applied to the plant is calibrated as a type of preset type of pesticide; If the proportion of the insect pest infection area of ​​the plant is greater than the preset proportion, the preset type of pesticide applied to the plant is obtained, indicating that the pest control effect of the preset type of pesticide does not meet the standard, and the preset type of pesticide applied to the plant is calibrated as a type II preset type pesticide.

5. The method for identifying and counting field pests based on image processing according to claim 3, characterized in that: According to the pest infection sub-region images of each plant, the characteristics of the pests infected by each plant are analyzed to obtain the pest types of the pests infected by each plant, specifically: If the similarity between the sub-region image and a feature image on a certain atlas sub-node is greater than a preset similarity threshold, then the index label marked on the atlas sub-node corresponding to the similarity greater than the preset similarity threshold is obtained; Determine the type of pests present in the corresponding pest-infected sub-region image according to the index labels marked on the graph sub-nodes corresponding to the graph sub-nodes whose similarity is greater than a preset similarity threshold; By analogy, the pest types existing in all pest-infected sub-region images of each plant are integrated to obtain the pest types of the pests infected in each plant.

6. The method for identifying field pests based on image processing according to claim 1, characterized in that: According to the pest infection sub-region images of each plant, the pests infected by each plant are counted and analyzed to obtain the number of pests infected by each plant and the pest images of each pest, specifically: Acquire the pest-infected sub-region image of each plant, and perform grayscale, Gaussian filtering and contrast enhancement processing on the pest-infected sub-region image; The Prewitt operator algorithm is used to calculate the gradient value of each pixel in the pest-infected sub-region image, and the pixel with the largest gradient value is marked as a seed point; Preset the size of the domain range, and calculate the gray value difference between the seed point and the adjacent pixel points within the preset domain range; If the grayscale value difference between a certain neighboring pixel and the seed point is greater than a preset difference threshold, the neighboring pixel is marked as an auxiliary pixel, and the auxiliary pixel is added to the area to which the seed point belongs; If the grayscale value difference between a certain neighboring pixel point and the seed point is not greater than a preset difference threshold, the neighboring pixel point is used as a new seed point; Repeat the above steps to add each pixel point to the area to which the seed point or the new seed point belongs, and iterate continuously until no new pixel point can be added to the area to which the seed point or the new seed point belongs, thereby obtaining a plurality of seed areas; Acquire image information of each seed region to separate a single pest in the pest-infected sub-region image from the image background, and obtain pest images of each pest in each plant; The regional quantity values ​​of the seed regions are counted, and the regional quantity values ​​of the seed regions are used as the number of pests infected in the corresponding plants.

7. The method for identifying field pests based on image processing according to claim 1, characterized in that: The camera device comprises: Camera, responsible for collecting images of plants in the field; Adjust the support module, which is responsible for fixing the camera and adjusting the camera shooting angle; A data storage module for storing image data; The power supply module provides power for the entire system.

8. A field pest identification and statistics system based on image processing, characterized in that: The field pest identification and statistics system based on image processing includes a memory and a processor. The memory stores a field pest identification and statistics method program based on image processing. When the field pest identification and statistics method program based on image processing is executed by the processor, the steps of the field pest identification and statistics method based on image processing as described in any one of claims 1 to 7 are implemented.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a field pest identification and statistical method program based on image processing. When the field pest identification and statistical method program based on image processing is executed by a processor, the steps of the field pest identification and statistical method based on image processing as described in any one of claims 1 to 7 are implemented.

Images