An intelligent sampling detection and analysis method for crop pests and diseases

Abstract

The present invention belongs to the technical field of pest and disease sampling detection and analysis. Specifically, an intelligent sampling detection and analysis method for crop pests and diseases is provided. The method includes: retrieving the growth points of abnormal plants by using a drone equipped with a spectral device and a positioning device, and dividing sampling areas corresponding to different abnormal categories; determining the sampling quantity within the area by analyzing the air environment and soil environment of different sampling areas, determining the sampling positions within the area by dividing the abnormal degree levels of the growth points of abnormal plants, generating a sampling environment information set, and importing the data as a training model; at the same time, generating a biochemical feature set of the sampled leaves through biochemical determination, using it as exported data, establishing a training model, and reflecting the mapping relationship from leaf characteristics and environmental data to biochemical characteristics. This method realizes the dynamic adjustment of the sampling strategy and improves the scientificity and representativeness of crop detection and analysis.

Description

Technical Field

[0001] The present invention belongs to the technical field of pest sampling detection and analysis, and relates to an intelligent sampling detection and analysis method for crop pests and diseases. Background Art

[0002] Crop pests and diseases are important factors affecting agricultural production. Traditional pest detection methods mainly rely on manual inspections and sampling, which have problems such as low efficiency, small coverage, and insufficient accuracy. With the development of drone technology and spectral analysis technology, modern agriculture has gradually introduced intelligent detection means. In the prior art, drones equipped with spectral devices can quickly scan farmland, identify abnormal reflection bands of plants, and thus locate pest and disease areas.

[0003] On this basis, there are many problems in the sampling link of traditional detection methods. For example, the determination of sampling positions and quantities often lacks a scientific basis and has a large randomness, and cannot comprehensively and accurately reflect the true distribution of crop pests and diseases in the farmland. This may lead to deviations in detection results, and further affect the formulation and implementation effects of subsequent control measures. Therefore, it is of great practical significance and application value to develop an intelligent sampling detection and analysis method for crop pests and diseases.

[0004] In the prior art, there are already some related solutions involving crop pest and disease detection, analysis, etc. For example, the patent with the Chinese patent publication number CN106338492A discloses a crop census system based on a small unmanned aerial vehicle (UAV) - borne imaging spectrometer, which includes: a small UAV platform and a mission system, and ground support equipment; the mission system includes: a measurement and control module, a GPS module, a small near - infrared imaging spectrometer, a data transmission module, and installation accessories; the crop census system obtains a target spectral image cube in a push - broom mode; the ground support equipment is responsible for completing the measurement and control of the small UAV and the mission system and the reception task of observation data. It can realize the census work for field crops, obtain the near - infrared spectral image cube data of the target, provide rich infrared images and spectral information, and has extremely high detection sensitivity and a relatively perfect evaluation system.

[0005] Compared with the above - mentioned solution, the following limitations exist: only generally conducting a census of farmland crops to obtain spectral image cube data, without accurately positioning and classifying different abnormal plants. In the detection and analysis of crop pests and diseases, the influence of air environment and soil environment on the occurrence and development of pests and diseases is not comprehensively considered, making it difficult to fully understand the causes and backgrounds of pest and disease occurrences, and thus not conducive to in - depth analysis of pest and disease problems and the formulation of effective control strategies, and the value of multi - source data cannot be fully utilized, resulting in deficiencies in the depth and breadth of crop pest and disease analysis. Summary of the Invention

[0006] In view of this, to solve the problems raised in the above background technology, a method for intelligent sampling detection and analysis of crop pests and diseases is proposed.

[0007] The object of the present invention can be achieved by the following technical solutions: The present invention provides a method for intelligent sampling detection and analysis of crop pests and diseases, including the following steps: Step 1: Use a drone to carry a spectral device and a positioning device, and screen out the growth points of plants with abnormal leaf surface reflection bands through spectral analysis, and mark the point cloud information of each abnormal plant growth point, including point cloud coordinates, abnormal categories, and degrees of abnormality.

[0008] Step 2: After clustering and dividing the point cloud coordinates of each abnormal plant growth point, mark each pest and disease area corresponding to each abnormal category, and detect the sampling compensation weights of each pest and disease area corresponding to each abnormal category affected by the air environment and soil environment through the deployment of sensor nodes, and determine the sampling quantity of each pest and disease area corresponding to each abnormal category.

[0009] Step 3: Divide the degree of abnormality level according to the sampling quantity, and determine each sampling position in each pest and disease area corresponding to each abnormal category.

[0010] Step 4: Use a drone to carry a sampling device to perform puncture sampling at each sampling position in each pest and disease area corresponding to each abnormal category, and temporarily store and mark the sampling leaf information according to the address range.

[0011] Step 5: After exporting the sampling leaves, perform secondary image analysis and determination of physiological and biochemical indexes before physiological and biochemical experiments to generate the mapping relationship between leaf characteristics and environmental data.

[0012] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) The present invention synchronously obtains soil dynamic environment data and air dynamic environment data, constructs a dual-source environment compensation weight for integrated analysis, and explores the potential relationship between environmental data and leaf characteristics. This intelligent analysis method can more deeply understand the occurrence mechanism and rules of crop pests and diseases, more accurately identify the ecological heterogeneity of pest and disease-related areas, improve the scientificity and representativeness of sampling, and thus provide more effective decision-making support for pest control.

[0013] (2) The present invention dynamically adjusts the sampling strategy by differentially determining the sampling quantity of different sampling areas in the farmland and optimizing the sampling positions within the area according to the degree of abnormality level of the abnormal plant growth points, avoiding the limitations of traditional fixed sampling methods. Description of the Drawings

[0014] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0015] Figure 1 It is a schematic flowchart of the implementation steps of the method of the present invention.

[0016] Figure 2 It is a schematic diagram of the construction of the space coordinate system of the present invention.

[0017] Figure 3 It is a structural display diagram of the sampling device carried by the drone of the present invention.

[0018] Reference numerals: 1, reference plane; 2, space height plane; 3, sampling arm member; 4, puncture cylinder member; 5, guide rail member; 6, memory. Detailed implementation manners

[0019] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0020] Please refer to Figure 1 As shown, the present invention provides an intelligent sampling detection and analysis method for crop pests and diseases. The method includes the following steps: step1, use a drone to carry a spectral device and a positioning device, and screen out the growth points of plants with abnormal leaf surface reflection bands through spectral analysis, and mark the point cloud information of each abnormal plant growth point, including point cloud coordinates, abnormal categories, and degrees of abnormality.

[0021] In the implementation manner, the screening out of the growth points of plants with abnormal leaf surface reflection bands through spectral analysis and marking the point cloud information of each abnormal plant growth point includes: constructing a spectral feature band adaptation library corresponding to each abnormal category of crops, and the abnormal categories include each growth defect category and each disease defect category. Among them, the growth defect category refers to that the dimensions such as the height and thickness of crop plants do not conform to the standard dimensions in this growth cycle, and the disease defect category refers to that the crop plants have leaf damage and bacterial erosion due to being eaten by insects or generating bacteria.

[0022] Specifically, the spectral feature band adaptation library for different categories of crops includes the feature bands of each growth defect category and the feature bands of each disease defect category. Among them, each growth defect category includes the plant height defect category, the thickness defect category, etc., and each disease defect category includes the curling defect category, the withering defect category, the rust disease defect category, etc.

[0023] Please refer to Figure 2 As shown, according to the difference in spectral reflection characteristics, identify whether the reflection band of the crop leaf belongs to the spectral feature band corresponding to the abnormal category in the adaptation library. Use the positioning device to count the point cloud positions of the crop leaves belonging to the spectral feature band corresponding to the abnormal category, generate each point cloud coordinate, which is defined as the point cloud coordinate of each abnormal plant growth point, and mark the abnormal category of each abnormal plant growth point. The point cloud coordinate takes the ground plane as the reference plane (xy coordinate system), and any vertical plane of the reference plane is the spatial height plane (yz coordinate system) to construct an xyz space coordinate system.

[0024] By extracting the spectral feature bands corresponding to each abnormal plant growth point, determine the degree of abnormality of each abnormal plant growth point. Specifically, the spectral data of the healthy plant growth point can be selected as the reference standard, and the spectral difference between the abnormal plant and the healthy plant in the feature band can be compared to evaluate the degree of abnormality. For example, calculate the difference in spectral reflectance in the feature band as an index to measure the degree of abnormality.

[0025] It should be noted that the spectral device is an instrument used to measure the distribution of light intensity with wavelength. Its principle mainly involves processes such as light dispersion, detection and conversion of optical signals. There are obvious differences in spectral reflectance between the leaves of normal plants and those suffering from diseases and pests or having growth defects. For example, the adaptation library records the typical spectral reflection characteristics of healthy wheat leaves in the visible light band (400 - 760nm) and the near-infrared band (760 - 1100nm), as well as the spectral feature band ranges under different abnormal conditions (abnormal sub-items) such as being infected with rust disease and lacking water. Use the spectral device to measure the spectrum of an actual wheat leaf and obtain the reflectance data at each wavelength. It is found that the reflectance of this wheat leaf at about 550nm is significantly lower than that of the healthy wheat leaf in the adaptation library, and it is consistent with the low reflectance characteristics in the 530 - 570nm band of the spectral characteristics of wheat infected with rust disease recorded in the adaptation library. At the same time, in the near-infrared band of 800 - 850nm, the reflectance also shows an abnormal increase, which also matches the spectral characteristics after wheat is infected with rust disease. Then, it can be judged that the reflection band of this wheat leaf belongs to the spectral feature band corresponding to the rust disease defect category in the adaptation library, thus determining that this wheat leaf belongs to an abnormal plant growth point, and its abnormal category is the rust disease defect category.

[0026] In addition, by collecting hyperspectral images, hyperspectral devices can not only extract spectral information but also spatial information. For example, if the thickness of plants is different, their surface texture features will vary, and these texture features will be reflected in hyperspectral images. The specific method is to obtain the hyperspectral images of plants, use image analysis technology to analyze the plant textures in the images, extract texture feature parameters such as gray-level co-occurrence matrix and local binary pattern, and judge whether the thickness of the plants is normal and identify whether there are growth category corresponding thickness defect categories through the analysis and comparison of these texture feature parameters.

[0027] Step 2: After clustering and dividing the point cloud coordinates of the growth points of different sample plants, mark each pest and disease area corresponding to each different sample category, and determine the sampling compensation weights of each pest and disease area corresponding to each different sample category affected by the air environment and soil environment by deploying sensor nodes, and determine the sampling quantity of each pest and disease area corresponding to each different sample category.

[0028] In the implementation manner, the content of marking each pest and disease area corresponding to each different sample category after clustering and dividing the point cloud coordinates of the growth points of different sample plants includes: obtaining the growth points of different sample plants corresponding to each different sample category by aggregating the growth points of different sample plants of the same different sample category.

[0029] Circle all the growth points of different sample plants corresponding to each different sample category into spatial regions according to the point cloud coordinates, and locate the geometric center point of each spatial region as the reference point of each different sample category.

[0030] Calculate the Euclidean distance between each reference point and the growth points of different sample plants, and regard the points whose Euclidean distance exceeds a certain multiple of the mean value as noise points and remove them.

[0031] Use a clustering algorithm to perform clustering analysis on the denoised points to form each cluster point cluster corresponding to each different sample category, and then define the region circled by each cluster point cluster corresponding to each different sample category as each pest and disease area corresponding to each different sample category.

[0032] Through clustering analysis, the growth points of different sample plants of the same different sample category form their respective corresponding cluster point clusters and pest and disease areas, clearly dividing the specific distribution areas of different pests and diseases in the farmland, providing a basis for subsequent targeted prevention and control.

[0033] In a further embodiment, the sampling compensation weights of each pest and disease area corresponding to each different sample category affected by the air environment and soil environment are detected by deploying sensor nodes, and the sampling quantity of each pest and disease area corresponding to each different sample category is determined. The content includes: step2-1, counting the number of growth points of all abnormal plants corresponding to each different sample category and the number of point cloud coordinates in each pest and disease area corresponding to each different sample category, and dividing the number of point cloud coordinates in each pest and disease area corresponding to each different sample category by the number of growth points of all abnormal plants corresponding to each different sample category to obtain the sampling quantity proportion of each pest and disease area corresponding to each different sample category , represents the number of different sample categories, , represents the number of each pest and disease area, .

[0034] step2-2, by deploying a variety of sensor nodes, such as temperature sensors, humidity sensors, etc., periodically collect the dynamic environment data of the farmland to generate a historical environment distribution database of the farmland, that is, including the dynamic soil environment data and dynamic air environment data that normally appear in different areas of the farmland. The dynamic soil environment data includes variable parameters such as pesticide accumulation content and water accumulation content, and the dynamic air environment data includes variable parameters such as wind direction, wind speed, sunshine intensity, and rainfall intensity.

[0035] step2-3, determine the soil accumulation structure of each pest and disease area corresponding to each different sample category according to the dynamic soil environment data, and then analyze the first sampling compensation weight of each pest and disease area corresponding to each different sample category .

[0036] step2-4, determine the air accumulation structure of each pest and disease area corresponding to each different sample category according to the dynamic air environment data, and then analyze the second sampling compensation weight of each pest and disease area corresponding to each different sample category .

[0037] step2-5, obtain the total planned sampling quantity , and determine the sampling quantity of each pest and disease area corresponding to each different sample category , where the total planned sampling quantity is the quantity fitted and set by technicians according to sampling experience, and there may be a certain error in the actually determined sampling quantity.

[0038] The sampling quantity method of each pest and disease area corresponding to each different sample category is as follows: , represents rounding down, represents the comprehensive proportion weight of the sampling quantity of each pest and disease area corresponding to each abnormal sample category. The sampling quantity of each pest and disease area corresponding to each different sample category is as follows in the table:

[0039] Table 1: Sampling Quantity Examples for Different Samples Corresponding to Each Pest and Disease Area

[0040]

[0041] In a further embodiment, step 2-3 includes: performing inertia analysis on the soil dynamic environment data to generate corresponding soil accumulation structures and soil inertia data for different sample categories corresponding to each pest and disease area. The soil accumulation structure includes areas with pesticide irrigation component accumulation, areas with excessive plant illumination, areas with insufficient illumination, and areas with soil waterlogging. The soil inertia data includes the corresponding accumulation amounts in the areas with pesticide irrigation component accumulation, the corresponding excessive amounts in the areas with excessive plant illumination, the corresponding insufficient amounts in the areas with insufficient illumination, and the corresponding waterlogging amounts in the areas with soil waterlogging.

[0042] Specifically, the soil accumulation structure refers to sub-areas in different sample categories corresponding to each pest and disease area that have phenomena such as pesticide irrigation component accumulation, excessive plant illumination, insufficient illumination, and soil waterlogging in multiple stage periods. Thus, the generated soil inertia data is the average value of different variable parameters in the soil dynamic environment data over multiple stage periods.

[0043] After converting the variable parameters included in the soil inertia data into a dimensionless form through the normalization method, weights are assigned to the corresponding variable parameters of the dimensionless soil inertia data based on the entropy weight method, such as pesticide content 0.4, illumination 0.3, and waterlogging 0.3, to eliminate the influence of different index dimensions and magnitudes, making the data comparable.

[0044] After multiplying the weighted values with the corresponding variable parameters of the dimensionless soil inertia data one by one and accumulating them, the soil environment complexity for different sample categories corresponding to each pest and disease area is obtained.

[0045] The soil environment complexity is converted into a probability value ranging from 0 to 1 through the Sigmoid function, which is defined as a type of sampling compensation weight for different sample categories corresponding to each pest and disease area. The greater the soil environment complexity, the more diverse the farmland structure, and thus the higher the demand for the sampling quantity of crops.

[0046] Specifically, the mathematical expression of the Sigmoid function is: , where A is the independent variable, is the probability value, and e is the natural constant.

[0047] In a further embodiment, the content of step 2-4 includes: performing an inertia analysis on the air dynamic environment data to generate an air accumulation structure and air inertia data corresponding to each pest and disease area for each different sample category. The air accumulation structure refers to sub-regions that contain phenomena such as hurricanes, sunlight exposure, and rainfall in multiple stages for each pest and disease area corresponding to each different sample category. The air inertia data is the average value of different variable parameters in the air dynamic environment data over multiple stages.

[0048] Normalize, assign values, perform mapping multiplication, cumulative calculation, and transformation on the respective variable parameters of the air inertia data to generate a probability value for the transformation of the air environment complexity corresponding to each pest and disease area for each different sample category, similar to the above-mentioned soil environment complexity transformation method.

[0049] Obtain the farmland space compensation structure, determine the position distances of the farmland space compensation structure relative to each pest and disease area corresponding to each different sample category, and convert each position distance into an environmental gradient difference through a spatial interpolation algorithm. The space compensation structure includes tree shelter structures, nearby factory structures, etc.

[0050] The spatial interpolation algorithm, such as spline interpolation, processes the distance data to generate a continuous spatial surface, and then divides the environmental gradient differences according to different distance ranges to generate environmental gradient differences for different regions, making the results more in line with the actual geographical space distribution.

[0051] According to the probability value of the transformation of the air environment complexity corresponding to each pest and disease area for each different sample category and the environmental gradient difference , determine the two-category sampling compensation weights corresponding to each pest and disease area for each different sample category , which is , used to reflect the degree of deviation impact of the space compensation structure on the air environment complexity.

[0052] Specifically, obstacles in the farmland, such as trees and buildings, will cause uneven spatial distribution of environmental factors such as light, temperature, and humidity in the farmland, increasing spatial heterogeneity. To accurately reflect the impact of this heterogeneity on research objects such as pests and diseases, it is necessary to increase the sampling quantity to cover samples in areas with different degrees of obstruction. At the same time, appropriate temperature and humidity are important conditions for the breeding and spread of pests and diseases. Small changes in temperature and humidity may lead to significant differences in the occurrence degree of pests and diseases. To timely grasp the relationship between the occurrence of pests and diseases and temperature and humidity, it is necessary to increase the sampling frequency and quantity in complex environmental areas with large temperature and humidity changes.

[0053] The present invention synchronously acquires soil dynamic environment data and air dynamic data, constructs a dual-source environmental compensation weight for integrated analysis, and explores the potential relationship between environmental data and leaf characteristics. This intelligent analysis method can more deeply understand the occurrence mechanism and laws of crop pests and diseases, more accurately identify the ecological heterogeneity of pest-associated regions, improve the scientificity and representativeness of sampling, and thus provide more effective decision-making support for pest control.

[0054] Step 3: Divide the degree of abnormality into levels according to the sampling quantity, and determine each sampling position in each pest and disease region corresponding to each abnormality category.

[0055] In the implementation manner, the dividing the degree of abnormality into levels according to the sampling quantity and determining each sampling position in each pest and disease region corresponding to each abnormality category includes: obtaining the leaf area corresponding to the growth points of abnormal plants in each region, setting an area threshold, and removing the growth points of abnormal plants with leaf area lower than the area threshold.

[0056] Divide the degree of abnormality of all the growth points of abnormal plants after the removal operation in each pest and disease region corresponding to each abnormality category according to the sampling quantity. Specifically, the sampling quantity is equivalent to the quantity of abnormality level division, and the abnormality level can be divided into level 0, level 1, level 2, level 3... and the healthy leaves in the region are used as the initial level, that is, level 0 is used as the healthy leaf level, so as to obtain samples under different degrees of abnormality, and thus a series of continuous and gradient data can be obtained in subsequent biochemical assays.

[0057] Select one growth point of an abnormal plant at different degrees of abnormality in each region respectively, and statistically obtain each sampling position in each pest and disease region corresponding to each abnormality category. For example, if the sampling quantity of a certain abnormality category corresponding to a certain pest and disease region is 10, then set the healthy leaves as level 0, and then select one growth point of an abnormal plant with leaves at level 0, level 1, level 2... level 10 respectively, and obtain 10 sampling positions of this abnormality category corresponding to this pest and disease region.

[0058] In the above implementation manner, by obtaining the leaf integrity and setting a threshold to remove the growth points of abnormal plants with low integrity, it can ensure that the sample leaves used for biochemical assays have a certain integrity and stability in structure and function, and only the leaves with specific integrity can more accurately reflect their biochemical characteristics in the normal physiological state or under the influence of pests and diseases, avoiding abnormal fluctuations in biochemical indexes caused by severely damaged leaves and affecting the accuracy and reliability of the measurement results. Taking healthy leaves as level 0 and selecting the growth points of plants with different degrees of abnormality according to the level division, the samples cover various states from healthy to different pest and disease degrees, which matches the requirement of subsequent biochemical assays to study the impact of pests and diseases on plant physiological biochemistry, and can comprehensively analyze the changes in biochemical indexes in plants at different stages of pest and disease development.

[0059] The present invention determines the sampling quantity of different sampling areas in the farmland through differentiation, and optimizes the sampling positions within the area according to the degree levels of abnormalities of the growth points of abnormal plants, realizing the dynamic adjustment of the sampling strategy and avoiding the limitations of traditional fixed sampling methods.

[0060] Step 4: Use the sampling device carried by the drone to perform puncture sampling on each sampling position within each pest and disease area corresponding to each abnormal category, and temporarily store and mark the information of the sampled leaves according to the address range.

[0061] Please refer to Figure 3 As shown, in the embodiment, the sampling device includes a sampling arm member, a puncture cylinder member, a guide rail member, and a memory.

[0062] In a further embodiment, the use of the sampling device carried by the drone to perform puncture sampling on each sampling position within each pest and disease area corresponding to each abnormal category, and temporarily store and mark the information of the sampled leaves according to the address range, specifically includes: after sending the puncture cylinder member to each sampling position through the sampling arm member, using the puncture cylinder member to perform puncture sampling on the plant leaves at each sampling position to obtain the sampled leaves, and then sending the sampled leaves to the memory through the guide rail member. In the memory, the address ranges of the sampled leaves within each pest and disease area corresponding to each abnormal category are divided and marked, so that the sampled leaves in each same area form a leaf group, and the information of the corresponding sampled leaves at each sampling position is marked by the memory. The information of the sampled leaves includes environmental information and growth defect information.

[0063] The environmental information includes soil dynamic environment data and air dynamic environment data, and the growth defect information includes the abnormal category and the degree of abnormality.

[0064] Step 5: After exporting the sampled leaves, perform secondary image analysis and determination of physiological and biochemical indexes before the physiological and biochemical experiment to generate the mapping relationship between leaf characteristics and environmental data.

[0065] In the embodiment, after exporting the sampled leaves, performing secondary image analysis and determination of physiological and biochemical indexes before the physiological and biochemical experiment to generate the mapping relationship between leaf characteristics and environmental data includes: sequentially exporting the leaf groups in different regions according to the address range.

[0066] Detect the morphological characteristic data, color characteristic data, and texture characteristic data of each sampled leaf to generate the corresponding sampling feature set of the leaf group in each pest and disease area corresponding to each abnormal category.

[0067] The morphological characteristic data includes the leaf area, leaf shape, etc., the color characteristic data includes the color values related to the chlorophyll content, the color values related to the content of other pigments, and the texture characteristic data includes the surface roughness, vein characteristics, etc.

[0068] Physiological and biochemical indexes of each leaf group are measured to generate corresponding biochemical feature sets of the leaf groups corresponding to each pest and disease area for different sample categories. The biochemical feature sets include photosynthetic physiological indexes (such as chlorophyll content, photosynthetic gas exchange parameters, fluorescence parameters), nutrient indexes (such as nitrogen, phosphorus, potassium contents and trace element contents), and stress resistance-related indexes (such as antioxidant enzyme activity, content of osmotic adjustment substances).

[0069] The corresponding sampling feature set of each leaf group and the sampling leaf information are imported into the data as the training model, and the biochemical feature set is used as the exported data, so as to establish a mapping relationship from leaf features and environmental data to biochemical features in the training model.

[0070] In the above embodiment, by establishing the training model, it is helpful to deeply understand the growth and development mechanism of crops, reveal the internal relationship between leaf features, environmental factors and biochemical features, provide a theoretical basis for agricultural scientific research, and promote the development of agricultural basic research. For example, based on the training model, it can be learned that under a specific environment, for leaves with certain color and texture features, what range their chlorophyll content usually lies in.

[0071] The above content is only an example and illustration of the concept of the present invention. Those skilled in the art of the present technology can make various modifications or supplements to the described specific embodiments or use similar ways to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, they should fall within the protection scope of the present invention.

Claims

1. A method for intelligent sampling, detection and analysis of crop pests and diseases, characterized in that: The method comprises the following steps: Step 1: Use drones equipped with spectral equipment and positioning equipment to screen out abnormal plant growth points in the leaf surface reflection band through spectral analysis, and mark the point cloud information of each abnormal plant growth point, including point cloud coordinates, abnormality category and abnormality degree; Step 2, after clustering the point cloud coordinates of the growth points of each abnormal plant, mark the pest and disease areas corresponding to each abnormal category, and deploy sensor nodes to detect the sampling compensation weights of each pest and disease area corresponding to each abnormal category affected by the air environment and soil environment, and determine the sampling quantity of each pest and disease area corresponding to each abnormal category, including: counting the number of point cloud coordinates of each abnormal category of plant growth points and pest and disease areas to generate the sampling quantity ratio, deploying sensors to collect farmland dynamic environment data to form a database, and determining the first and second category sampling compensation weights respectively according to the soil and air dynamic environment data, and then combining the total planned sampling volume to determine the final sampling quantity of each pest and disease area; Step 3: Classify the degree of abnormality according to the number of samples, and determine the sampling locations in each pest and disease area corresponding to each abnormality category; Step 4: Use the drone equipped with sampling equipment to puncture and sample each sampling location in each pest and disease area of ​​each abnormal category, and temporarily store and mark the sampled leaf information according to the address range; Step 5. After exporting the sampled leaves, perform secondary image analysis and physiological and biochemical index determination before physiological and biochemical experiments to generate a mapping relationship between leaf characteristics and environmental data.

2. The method for intelligent sampling, detection and analysis of crop pests and diseases according to claim 1, characterized in that: The method of screening out abnormal plant growth points in the leaf surface reflection band through spectral analysis and marking point cloud information of each abnormal plant growth point includes: Construct a library of spectral characteristic band adaptations for different types of crops; According to the difference in spectral reflectance characteristics, it is identified whether the reflection band of the crop leaves belongs to the corresponding spectral characteristic band of the abnormal category in the adaptation library, and the positioning equipment is used to perform point cloud position statistics on the crop leaves belonging to the corresponding spectral characteristic band of the abnormal category, and each point cloud coordinate is generated, which is defined as the point cloud coordinate of each abnormal plant growth point, and the abnormal category of each abnormal plant growth point is marked; By extracting the corresponding spectral characteristic bands of the growth points of various abnormal plants, the abnormal degree of the growth points of various abnormal plants can be determined.

3. The intelligent sampling, detection and analysis method for crop pests and diseases according to claim 2 is characterized in that: After clustering the point cloud coordinates of the growth points of each abnormal plant, the pest and disease areas corresponding to each abnormal category are marked, including: By gathering the growth points of different plants of the same different category, the growth points of different plants corresponding to different categories are obtained; According to the point cloud coordinates, all the abnormal plant growth points corresponding to each abnormal category are circled into spatial regions, and the geometric center point of each spatial region is located as the reference point of each abnormal category; Calculate the Euclidean distance between each reference point and the growth point of the abnormal plant, and remove the points whose Euclidean distance exceeds a certain multiple of the mean as noise points; A clustering algorithm is used to perform cluster analysis on the denoised points to form clusters of cluster points corresponding to each abnormal category, and then the areas circled by the cluster points corresponding to each abnormal category are defined as the pest and disease areas corresponding to each abnormal category.

4. The method for intelligent sampling, detection and analysis of crop pests and diseases according to claim 1, characterized in that: Determine the number of samples to be taken from each pest and disease area for each abnormal category, including: Step 2-1, count the number of all abnormal plant growth points corresponding to each abnormal category and the number of point cloud coordinates in each pest and disease area corresponding to each abnormal category, and generate the proportion of the number of samples corresponding to each pest and disease area for each abnormal category; Step 2-2, by deploying a variety of sensor nodes, the dynamic environmental data of farmland is collected periodically to generate a farmland historical environmental distribution database, which includes the soil dynamic environmental data and air dynamic environmental data that normally appear in different areas of the farmland; Step 2-3, determine the soil accumulation structure of each abnormal category corresponding to each pest and disease area based on soil dynamic environmental data, and then analyze the sampling compensation weight of each abnormal category corresponding to each pest and disease area; Step 2-4: Determine the air accumulation structure of each abnormal category corresponding to each pest and disease area based on the air dynamic environment data, and then analyze the second-class sampling compensation weight of each abnormal category corresponding to each pest and disease area; Step 2-5, obtain the total planned sampling quantity and determine the sampling quantity for each anomaly category corresponding to each pest and disease area.

5. The method for intelligent sampling, detection and analysis of crop pests and diseases according to claim 4, characterized in that: The step 2-3 includes: Performing habit analysis on soil dynamic environment data to generate soil accumulation structure and soil habit data corresponding to each pest and disease area of ​​each abnormal category, wherein the soil accumulation structure includes pesticide irrigation component accumulation area, plant excessive light area and light deficiency area, and soil water accumulation area, and the soil habit data includes the corresponding accumulation amount of pesticide irrigation component accumulation area, the corresponding excess amount of plant excessive light area and the corresponding deficiency amount of light deficiency area, and the corresponding water accumulation amount of soil water accumulation area; After the variable parameters contained in the soil habit data are converted into dimensionless form by normalization method, the corresponding variable parameters of the dimensionless soil habit data are weighted based on the entropy weight method; After mapping and multiplying the weighted values ​​and the corresponding variable parameters of the dimensionless soil habit data one by one, the soil environmental complexity of each disease and insect pest area corresponding to each abnormal category is accumulated; The soil environmental complexity was converted into a probability value ranging from 0 to 1 through the Sigmoid function, which was defined as a sampling compensation weight for each anomaly category corresponding to each pest and disease area.

6. The method for intelligent sampling, detection and analysis of crop pests and diseases according to claim 5, characterized in that: Step 2-4 includes: Conduct a trend analysis on the dynamic air environment data to generate the air accumulation structure and air trend data corresponding to each pest and disease area of ​​each category; Normalize, assign values, map products, accumulate and transform the corresponding variable parameters of the air habitus data to generate the probability value of the air environment complexity transformation corresponding to each pest and disease area for each abnormal category; Obtain the farmland spatial compensation structure, determine the location distance of the farmland spatial compensation structure relative to each pest and disease area corresponding to each heterogeneous category, and convert each location distance into an environmental gradient difference through a spatial interpolation algorithm; According to the probability value of the transformation of the air environment complexity of each anomaly category corresponding to each pest and disease area and the difference in environmental gradient, the second-category sampling compensation weight of each anomaly category corresponding to each pest and disease area is determined.

7. The method for intelligent sampling, detection and analysis of crop pests and diseases according to claim 1, characterized in that: The abnormality degree is divided into grades according to the number of samples, and the sampling locations in each pest and disease area corresponding to each abnormality category are determined, including: Obtain the corresponding leaf areas of abnormal plant growth points in each region, set an area threshold, and remove abnormal plant growth points whose leaf areas are lower than the area threshold; According to the number of samples, all abnormal plant growth points after culling in each pest and disease area of ​​each abnormal category are classified into abnormal degree levels; An abnormal plant growth point with different abnormality levels in each area is selected respectively, and the sampling positions of each abnormality category corresponding to each pest and disease area are statistically obtained.

8. The intelligent sampling, detection and analysis method for crop pests and diseases according to claim 1, characterized in that: The sampling device comprises a sampling arm component, a puncture cylinder component, a guide rail component and a storage device.

9. The intelligent sampling, detection and analysis method for crop pests and diseases according to claim 8, characterized in that: The method uses a sampling device mounted on an unmanned aerial vehicle to perform puncture sampling on each sampling position in each pest and disease area corresponding to each abnormal category, and temporarily stores and marks the sampled leaf information according to the address range. The specific content is: after the puncture tube component is sent to each sampling position by the sampling arm component, the puncture tube component is used to puncture and sample the plant leaves at each sampling position to obtain the sampled leaves, and then the sampled leaves are sent to the memory through the guide rail component. In the memory, the address range of the sampled leaves in each pest and disease area corresponding to each abnormal category is divided and marked, so that the sampled leaves in each same area form a leaf group, and the corresponding sampled leaf information of each sampling position is marked through the memory, and the sampled leaf information includes environmental information and growth defect information.

10. The method for intelligent sampling, detection and analysis of crop pests and diseases according to claim 9, characterized in that: After the sampled leaves are exported, secondary image analysis and physiological and biochemical index determination are performed before the physiological and biochemical experiment to generate a mapping relationship between leaf characteristics and environmental data, including: Export leaf groups in different regions sequentially by address range; Detect the morphological feature data, color feature data and texture feature data of each sampled leaf, and generate a corresponding sampling feature set of leaf groups corresponding to each pest and disease area for each abnormal category; The physiological and biochemical indexes of each leaf group were measured to generate the corresponding biochemical characteristic set of leaf groups of each heterogeneous category corresponding to each pest and disease area; The corresponding sampling feature set of each leaf group and the sampled leaf information are used as the training model import data, and the biochemical feature set is used as the export data, and then a mapping relationship from leaf features and environmental data to biochemical features is established in the training model.

Images