Astragalus sinicus seedling emergence recognition statistical method and device

Abstract

The invention discloses an astragalus sinicus seedling emergence recognition statistical method and device, and the method comprises the steps: collecting plant feature images of different periods after the seedling emergence of astragalus sinicus, and building an image database; performing image acquisition on seedlings after field astragalus sinicus emerges, comparing the acquired seedling images with the plant feature images in the image database, calculating the proportion of the field plants with astragalus sinicus emerging plants and the plant class group composition, and obtaining the emergence rate and the distribution condition of astragalus sinicus; and according to the emergence rate and the distribution condition, quantitative seed reseeding and quantitative use and spraying of a special herbicide are carried out. According to the seedling emergence recognition statistical method, the labor cost is saved, the efficiency is improved, meanwhile, the accuracy is greatly improved, the field astragalus sinicus seedling emergence condition can be rapidly known, and corresponding remedial measures such as reseeding or spraying of special herbicide for weed protection can be conveniently taken.

Description

Technical Field

[0001] The invention belongs to the field of image recognition, and in particular relates to a method and device for statistically identifying emergence of Chinese milk vetch seedlings. Background Art

[0002] Now with the widespread application of green manure, milk vetch plays an important role in soil improvement, rice field rotation and other scenarios, and the planting area is getting larger and larger. It is generally sown in autumn. After sowing, it is necessary to timely understand and grasp the emergence of field planting in detail so as to re-sow as soon as possible. However, at present, only manual on-site inspections can be used. Since milk vetch plants are very small in the seedling stage and difficult to distinguish from weeds, it is necessary to bend over and carefully identify the specific emergence rate. In addition, the fields after sowing are kept at a certain humidity in order to germinate as soon as possible, which makes walking inconvenient. If it is a large-scale planting field, it is difficult to ensure the uniform emergence of milk vetch, thereby reducing its ability to improve soil and help increase grain production. Summary of the Invention

[0003] To solve the above technical problems, the present invention provides a method and device for identifying and counting the emergence of Chinese milk vetch seedlings. A method for identifying and counting the emergence of Chinese milk vetch seedlings includes:

[0004] Collect characteristic images of Chinese milk vetch plants at different stages after seedling emergence and establish an image database;

[0005] Collect images of the milk vetch seedlings after they emerge in the field, compare the collected seedling images with the plant feature images in the image database, calculate the proportion of the milk vetch seedlings in the field plants and the plant group composition, and obtain the emergence rate and distribution of the milk vetch;

[0006] According to the emergence rate and distribution, quantitative seed reseeding and quantitative use and spraying of special herbicides are carried out.

[0007] Preferably, the process of collecting characteristic images of Chinese milk vetch plants at different stages after emergence and establishing an image database includes:

[0008] The cameras were set at different positions and angles to collect images of the milk vetch seedlings in the field from multiple angles and at different time periods after emergence, thus obtaining all-round plant characteristic images of the milk vetch seedlings at different stages after emergence.

[0009] Image preprocessing technology was used to denoise and enhance the characteristic images of plants at different stages to obtain the morphological characteristics of Chinese milk vetch plants.

[0010] Preferably, the process of using image preprocessing technology to perform denoising and enhancement processing on plant feature images at different stages to obtain the morphological characteristics of the milk vetch plant includes:

[0011] The median filter algorithm is used to denoise the plant feature image to obtain a denoised image dataset;

[0012] Performing image enhancement processing on the denoised image dataset to obtain an enhanced image dataset;

[0013] Based on the enhanced image dataset, the K-means clustering algorithm is used to divide the Chinese milk vetch planting area into regions to obtain divided image subsets;

[0014] Extract plant features from the divided image subsets to obtain a feature vector set; the feature vector set includes leaf morphology, plant height, leaf length-to-width ratio, leaf color depth, edge smoothness, serrations, hairiness, and stem-to-leaf ratio;

[0015] Classifying the feature vector set by a random forest algorithm to determine the growth stage of the Chinese milk vetch and obtain a growth stage label set;

[0016] According to the growth stage label set, the changing trends of plant characteristics at different growth stages are analyzed to obtain a characteristic change data set.

[0017] Preferably, the process of performing image enhancement processing on the denoised image dataset to obtain the enhanced image dataset includes:

[0018] According to the denoised image dataset, an enhancement operation is used to adjust the contrast and brightness of the image to highlight the morphological characteristics of the milk vetch plant, thereby obtaining a morphological feature image set;

[0019] For the morphological feature image set, edge detection technology is applied to extract the morphological features of the milk vetch plant and determine the feature-significant area;

[0020] If the clarity of the feature-salient area is lower than a preset threshold, performing local sharpening processing on the morphological feature image set to obtain a sharpened image set;

[0021] According to the sharpened image set, the morphological features of the Chinese milk vetch plant are identified using a pre-established classification model to determine the feature category;

[0022] According to the distribution of feature categories, a cluster analysis method is used to group the sharpened image set to obtain a clear image set;

[0023] If noise interference exists in the clear image set, a secondary denoising process is performed on the grouped images to determine a final enhanced image data set.

[0024] Preferably, the process of calculating the proportion of Chinese milk vetch plants that have emerged in the field and the composition of the plant groups includes:

[0025] The morphological characteristics of Chinese milk vetch plants were extracted and classified using a convolutional neural network to identify individual plants at different growth stages and obtain classified feature label data.

[0026] The classified feature label data is combined with a field grid division method to divide the planting area into sub-areas, and the distribution density of the milk vetch plants in each sub-area is calculated to obtain regional distribution ratio information;

[0027] If the distribution density of a sub-region in the regionalized distribution ratio information is lower than a preset threshold, a marking mechanism for the low-density region is triggered to generate marked distribution abnormality data;

[0028] According to the marked distribution abnormality data, image comparison technology is used to match and analyze the plant characteristics in the abnormal area with a pre-established standard image database of milk vetch, to determine the growth abnormality and obtain a classification result of the abnormality cause.

[0029] Preferably, the process of quantitative seed reseeding and quantitative use and spraying of special herbicides according to the emergence rate and distribution includes:

[0030] Automatically search for areas that need to be reseeded based on the emergence rate and distribution of seedlings;

[0031] According to the emergence of milk vetch in different areas, quantitative sowing and reseeding are carried out according to the set seed quantity;

[0032] At the same time, based on the distribution information of field plant groups, the usage amount of special herbicides is determined, and the special herbicides are quantitatively used and sprayed according to the set usage amount.

[0033] The present invention also provides a device for identifying and counting the emergence of Chinese milk vetch seedlings, comprising:

[0034] A data storage system for collecting characteristic images of Chinese milk vetch plants at different stages after seedling emergence and establishing an image database;

[0035] Plant information collection and identification system, used to collect images of Chinese milk vetch seedlings after they emerge in the field;

[0036] A data comparison and analysis system is used to compare the collected seedling images with the plant feature images in the image database, calculate the proportion of Chinese milk vetch seedlings in the field plants and the composition of plant groups, and obtain the emergence rate and distribution of Chinese milk vetch;

[0037] The material precision delivery system is used to carry out quantitative seed reseeding and quantitative use and spraying of special herbicides according to the emergence rate and distribution.

[0038] Preferably, the plant information acquisition and identification system includes an image acquisition module and an image processing module;

[0039] The image acquisition module is used to acquire images of the milk vetch seedlings in the field after emergence from multiple angles and different time periods by setting cameras at different positions and angles, thereby obtaining all-round plant feature images of the milk vetch seedlings at different stages after emergence in the field;

[0040] The image processing module is used to perform denoising and enhancement processing on plant feature images at different stages using image preprocessing technology to obtain the morphological characteristics of the milk vetch plant.

[0041] Preferably, the material precision delivery system includes a quantitative module and a spraying module;

[0042] The quantitative module is used to automatically search for the area that needs to be reseeded according to the emergence rate and distribution of the seedlings; according to the emergence of the milk vetch in different areas, quantitative sowing and reseeding are carried out according to the set seed quantity;

[0043] The spraying module is used to determine the usage amount of the special herbicide according to the distribution information of the plant groups in the field, and quantitatively use and spray the special herbicide according to the set usage amount.

[0044] Preferably, the device further comprises a data updating module and a user interaction module;

[0045] The data updating module is used to update and optimize the plant feature image database of the Chinese milk vetch at different stages after seedling emergence based on the actually collected image data;

[0046] The user interaction module is used to display the emergence rate and distribution of Chinese milk vetch, the amount of reseeding seeds, and the amount of herbicide used, and allows the user to operate and set.

[0047] Compared with the prior art, the present invention has the following advantages and technical effects:

[0048] The method of the present invention can quickly identify and count the emergence of milk vetch seedlings after large-scale sowing of milk vetch in the field, so as to achieve precise positioning and accurate judgment;

[0049] The method of the present invention can automatically generate a field distribution map of the emergence rate of the milk vetch after obtaining the emergence situation of the milk vetch in the field, calculate the amount of seeds that need to be re-sown according to the set standard, and then re-enter the field to perform quantitative automatic re-sowing according to the previous emergence rate statistical map.

[0050] The seedling emergence identification statistical method of the present invention saves labor costs and improves efficiency, while greatly enhancing accuracy. It can quickly understand the emergence of Chinese milk vetch in the field and facilitate the formulation of corresponding remedial measures, such as re-sowing seeds or spraying special herbicides for weed protection. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The accompanying drawings, which constitute part of this application, are intended to provide a further understanding of this application. The exemplary embodiments and descriptions of this application are intended to explain this application and do not constitute an improper limitation on this application. In the accompanying drawings:

[0052] Figure 1 Schematic diagram of a method flow in an embodiment of the present invention. DETAILED DESCRIPTION

[0053] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0054] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0055] Example 1

[0056] like Figure 1 As shown, this embodiment provides a method for identifying and statistically analyzing the emergence of Chinese milk vetch seedlings, including:

[0057] Collect characteristic images of Chinese milk vetch plants at different stages after seedling emergence and establish an image database;

[0058] Image collection was performed on the seedlings of Chinese milk vetch after they emerged in the field. The collected seedling images were compared with the plant feature images in the image database. The proportion of Chinese milk vetch seedlings in the field plants and the composition of plant groups were calculated to obtain the emergence rate and distribution of Chinese milk vetch.

[0059] According to the emergence rate and distribution, quantitative seed reseeding and quantitative use and spraying of special herbicides are carried out.

[0060] Furthermore, the process of collecting characteristic images of Chinese milk vetch plants at different stages after emergence and establishing an image database includes:

[0061] The cameras were set at different positions and angles to collect images of the milk vetch seedlings in the field from multiple angles and at different time periods after emergence, thus obtaining all-round plant characteristic images of the milk vetch seedlings at different stages after emergence.

[0062] Image preprocessing technology was used to denoise and enhance the characteristic images of plants at different stages to obtain the morphological characteristics of Chinese milk vetch plants.

[0063] Furthermore, image preprocessing technology is used to perform denoising and enhancement on the characteristic images of plants at different stages. The process of obtaining the morphological characteristics of the Chinese milk vetch plant includes:

[0064] The median filter algorithm is used to denoise the plant feature image to obtain a denoised image dataset;

[0065] Perform image enhancement processing on the denoised image dataset to obtain an enhanced image dataset;

[0066] Based on the enhanced image dataset, the K-means clustering algorithm was used to divide the Chinese milk vetch planting area and obtain the divided image subsets.

[0067] Extract plant features from the divided image subsets to obtain a feature vector set; the feature vector set includes leaf morphology, plant height, leaf length-to-width ratio, leaf color depth, edge smoothness, serrations, hairiness, and stem-to-leaf ratio;

[0068] The feature vector set is classified by random forest algorithm to determine the growth stage of Chinese milk vetch and obtain the growth stage label set;

[0069] According to the growth stage label set, the changing trends of plant characteristics in different growth stages are analyzed to obtain the feature change dataset.

[0070] Specifically, this embodiment uses high-resolution field image acquisition equipment to acquire raw image data from a Chinese milk vetch planting area, covering plant features at different growth stages. The specific implementation method for forming an initial image dataset can be integrated into a complete process. First, a high-resolution camera-equipped drone is used to conduct a grid-based flight scan at an altitude of 50 meters above the Chinese milk vetch planting area, with a resolution of 0.05 meters per pixel to ensure coverage of 100 hectares. Collection times are set during the budding, growth, and flowering stages of the Chinese milk vetch, with 5,000 images collected during each stage, forming a total raw image dataset of 15,000 images. Next, an image preprocessing algorithm is used to denoise the collected images. A Gaussian filter is used with a filter kernel size of 3x3 and a standard deviation of 1.5 to reduce the effects of noise caused by ambient lighting and equipment jitter. Grayscale histogram equalization is used to enhance image contrast, ensuring that plant features remain discernible at least 90% under different lighting conditions. Subsequently, deep learning models such as YOLOv5 were used to detect objects in the images. The model was trained to identify the characteristics of Chinese milk vetch plants at different growth stages. The training set accounted for 80%, the validation set accounted for 20%, and the number of iterations was set to 100. Detection accuracy reached over 85%, and plant density data for each growth stage was output. For example, the number of plants per square meter during the budding stage was 20, the number of plants per square meter during the active growth stage was 35, and the number of plants per square meter during the flowering stage was 50. Finally, the processed image data was integrated with the detection results to construct a structured database storing plant characteristic parameters such as leaf area, stem height, and number of flowers. A time series analysis algorithm was then used to predict the growth trend of Chinese milk vetch. Using the ARIMA model with parameters p = 2, d = 1, and q = 1, the plant density change over the next seven days was predicted with an error of less than 5%.

[0071] Furthermore, the denoised image dataset is subjected to image enhancement processing to obtain an enhanced image dataset, and the process includes:

[0072] Based on the denoised image dataset, the enhancement operation was used to adjust the contrast and brightness of the image to highlight the morphological characteristics of the milk vetch plant and obtain a morphological feature image set.

[0073] For the morphological feature image set, edge detection technology is applied to extract the morphological features of the milk vetch plant and determine the characteristic significant areas;

[0074] If the clarity of the feature-salient area is lower than a preset threshold, the morphological feature image set is locally sharpened to obtain a sharpened image set;

[0075] Based on the sharpened image set, the morphological features of the Chinese milk vetch plant are identified using a pre-established classification model to determine the feature category;

[0076] According to the distribution of feature categories, cluster analysis method is used to group the sharpened image set to obtain the clear image set;

[0077] If there is noise interference in the clear image set, the grouped images are subjected to secondary denoising processing to determine the final enhanced image data set.

[0078] Specifically, this embodiment performs denoising and enhancement processing on the image data set to highlight the morphological characteristics of the milk vetch plant. The specific implementation method is as follows: First, for image denoising, the Gaussian filtering algorithm is used to process the initial image, and the filter kernel size is set to 5x5 and the standard deviation is 1.5. By calculating the weighted average around each pixel point, the high-frequency noise in the image is effectively removed while retaining the edge information of the milk vetch plant. The analysis results show that the noise intensity is reduced from the initial 25.6 to 8.3, and the signal-to-noise ratio is improved by about 15%. Subsequently, in order to further enhance the image details, the adaptive histogram equalization (CLAHE) algorithm is applied, and the contrast limit parameter is set to 2.0 and the grid size is 8x8 to enhance the texture and contour features of the milk vetch plant. Especially under the condition of uneven lighting, the analysis shows that the contrast of the processed image is improved by about 30%, and the grayscale gradient value of the leaf edge of the plant is increased from the original 12.5 to 18.7, and the morphological characteristics are more significant. Next, to highlight the color characteristics of the Chinese milk vetch plant, HSV color space conversion was used to convert the image from RGB space to HSV space. The S channel (saturation) was stretched and adjusted from a linear mapping range of 0-255 to 20-235 to enhance the color distinction between the plant and the background. Analysis showed that the color difference between the plant area and the background increased from 15 to 28 after treatment. Finally, morphological operations (such as opening) were used to remove small area noise points. The structuring element was set to a 3x3 rectangular kernel and the number of iterations was set to 2 to ensure that the main morphology of the Chinese milk vetch plant in the image was not disturbed. Analysis results showed that the proportion of small noise points decreased from 5% to 0.5%, resulting in a clear image set.

[0079] Furthermore, in the enhancement operation of contrast and brightness adjustment, the histogram equalization technique is used to highlight the morphological characteristics of the milk vetch plant.

[0080] Specifically, for the images in the first image set, we can analyze their grayscale distribution, increase the brightness of the darker areas by 10%-15%, and appropriately reduce the contrast of the overly bright areas to ensure that the leaf edges and stem textures are clearly presented under different lighting environments. This approach helps to increase the accuracy of subsequent feature extraction.

[0081] For example, in the application of edge detection technology, this embodiment uses a classic edge detection operator to extract the morphological features of the milk vetch plant.

[0082] In one embodiment, for the morphological feature image set, the image may be first converted into a grayscale image, and then the boundary area between the leaf and stem may be identified by detecting the grayscale change of pixels, thereby determining the feature-significant area.

[0083] If it is found that the edge clarity of some areas is lower than a preset threshold, for example, lower than 80%, local sharpening processing is performed on the morphological feature image set to enhance edge details and obtain a sharpened image set.

[0084] For example, when using a classification model to identify morphological features, features such as the leaf shape and stem thickness of the asparagus can be categorized based on a pre-labeled data set.

[0085] For example, the features can be divided into three categories: seedling stage, growth stage and maturity stage. The model outputs the category distribution of plants in each image to provide a basis for subsequent analysis.

[0086] For example, in the cluster analysis grouping step, the images in the sharpened image set can be grouped into subsets of different growth stages according to the distribution of feature categories.

[0087] In one possible implementation, the image can be divided into two groups, high-density areas and low-density areas, according to plant density and morphological similarity, to obtain a set of clear images.

[0088] This grouping method helps to analyze the growth conditions of different areas in a targeted manner.

[0089] For example, in the secondary denoising process, if noise interference still exists in the clear image set, the local mean filtering method can be used to further process the grouped images.

[0090] For example, for images in high-density areas, a smaller filter window, such as 3x3 pixels, can be set to reduce noise while retaining details, ultimately determining a clear image set.

[0091] This multi-level processing method can effectively improve the availability of image data and provide reliable support for the precise management of milk vetch planting.

[0092] Furthermore, the process of calculating the proportion of Chinese milk vetch plants that have emerged in the field and the composition of plant groups includes:

[0093] The morphological characteristics of Chinese milk vetch plants were extracted and classified using a convolutional neural network to identify individual plants at different growth stages and obtain classified feature label data.

[0094] By using the classified feature label data and combining it with the field grid division method, the planting area is divided into sub-areas, and the distribution density of milk vetch plants in each sub-area is calculated to obtain regional distribution ratio information.

[0095] If the distribution density of a sub-region in the regionalized distribution ratio information is lower than the preset threshold, the marking mechanism of the low-density region is triggered to generate marked distribution abnormality data;

[0096] According to the marked distribution abnormality data, image comparison technology is used to match and analyze the plant characteristics in the abnormal area with the pre-established standard image database of Chinese milk vetch to determine the growth abnormality and obtain the classification result of the abnormal cause.

[0097] Furthermore, the morphological characteristics of the milk vetch plants are extracted and classified based on the convolutional neural network, and individual plants at different growth stages are identified. The process of obtaining the classified feature label data includes:

[0098] A convolutional neural network is used to extract morphological features and generate the first eigenvector;

[0099] Classifying the first eigenvector using a pre-trained network model to determine the growth stage of the individual plant and obtain a first classification result;

[0100] If the confidence level of the first classification result is lower than a preset threshold, extended image data is generated through data enhancement, and the second eigenvector is re-extracted;

[0101] A support vector machine is used to perform secondary classification on the second eigenvector to obtain a second classification result;

[0102] The second classification result is converted into a feature label through label mapping to generate label data;

[0103] In one possible implementation, when using a convolutional neural network to extract morphological features, the convolution operation scans the image using multiple layers of convolution kernels, capturing local features such as the edges and texture of the milk vetch plant, thereby generating a first feature vector. Assume that the network architecture incorporates multiple convolutional and pooling layers, gradually extracting feature information from low-level to high-level levels. This approach can effectively identify the plant's leaf shape and stem structure, providing key information for subsequent classification.

[0104] For example, to classify the first feature vector through a pre-trained network model to determine the growth stage, it is conceivable to use a deep learning model trained on a large-scale plant image dataset, input the feature vector into the model, and output the first classification result corresponding to the growth stage.

[0105] Specifically, the model may classify plants into different stages, such as seedling, growth, and maturity. This classification method can quickly identify the developmental status of the plant and provide a reference for subsequent analysis.

[0106] For example, if the confidence level of the first classification result falls below a preset threshold, data augmentation can be used to generate extended image data. For example, if the preset threshold is 0.8 and the confidence level of a classification result is only 0.6, the original image can be transformed by rotating, flipping, or adjusting the brightness to generate more diverse image data, allowing the second eigenvector to be re-extracted. This approach effectively increases data diversity and improves classification robustness.

[0107] For example, when using a support vector machine to perform a secondary classification on the second eigenvector, one can imagine constructing a hyperplane in a high-dimensional space to distinguish features at different growth stages and obtain a secondary classification result. This approach is particularly suitable for small datasets and can find the optimal boundary in the feature space, enhancing classification accuracy.

[0108] For example, when converting the second classification result into a feature label through label mapping, the classification result can be mapped one-to-one with the predefined growth stage label to generate labeled data. Assuming the classification result is the growth stage, it is mapped to the corresponding label number, which facilitates subsequent statistics and analysis.

[0109] For example, when extracting statistical information from label data to obtain growth stage distribution data, we can aggregate the labels of all individual plants and analyze the proportion of different growth stages. For example, if 100 plants are sampled, 30% are in the seedling stage, 50% are in the growth stage, and 20% are in the mature stage. This distribution data can intuitively reflect the development of the population and provide an important reference for subsequent research.

[0110] Furthermore, by using the classified feature label data and combining it with the field grid division method, the planting area is divided into sub-areas, and the distribution density of milk vetch plants in each sub-area is calculated. The process of obtaining regional distribution ratio information includes:

[0111] By combining classification data with feature labels, we can obtain preliminary distribution information of Chinese milk vetch plants in the planting area and determine the initial distribution characteristics.

[0112] According to the field grid and division method, the planting area is segmented to obtain the division results of multiple sub-areas;

[0113] The grid division results were used to collect statistics on the distribution of Chinese milk vetch plants in each sub-region and calculate the distribution density.

[0114] If the distribution density is within a preset threshold range, the sub-region segment is marked as a uniformly distributed region, and the uniformly distributed regional distribution data is obtained;

[0115] If the distribution density exceeds the preset threshold range, it is marked as a non-uniform distribution area and the distribution data of the non-uniform distribution area is recorded;

[0116] By combining regional distribution data with proportion information, the proportion of evenly distributed areas and unevenly distributed areas is calculated to determine the distribution proportion information of each area;

[0117] Based on the distribution ratio information and density calculation results, the distribution density comparison data of the sub-region segments is generated to determine the density difference characteristics between the sub-region segments;

[0118] After obtaining the density difference characteristics, the distribution density of Chinese milk vetch plants and the regional distribution data were combined to construct a distribution density distribution map to obtain the final regional distribution ratio information.

[0119] Specifically, by processing the classified feature label data and combining it with the field grid division method, the planting area is divided into multiple sub-areas, and the distribution density of milk vetch plants in each sub-area is calculated, and finally the regional distribution ratio information is obtained. First, assume that the feature label data of a milk vetch planting area with a total area of ​​10,000 square meters is obtained. This data includes the plant location coordinates and high-resolution image classification results, marking a total of 5,000 plant distribution points. Using the grid division algorithm, the entire area is divided into 100 sub-areas, each with an area of ​​100 square meters. The grid division adopts the uniform segmentation method, and the grid boundary is automatically generated by the geographic information system software to ensure that the boundary coordinates of each grid are accurate to 0.1 meters. Subsequently, based on the plant location coordinates in the feature label data, the number of plants in each sub-region is counted. For example, if there are 60 plants in sub-region A1 and 45 plants in sub-region A2, when calculating the distribution density, the density of sub-region A1 is 60 / 100 = 0.6 plants / square meter, and the density of sub-region A2 is 45 / 100 = 0.45 plants / square meter. Similarly, the density calculation is completed by traversing all sub-regions. Next, the density distribution characteristics of each sub-region are analyzed. Assuming that the density range is between 0.2 and 0.8 plants / square meter, a clustering algorithm (such as K-means) is used to divide the density into three categories: high, medium, and low. The distribution ratio information is obtained: high-density areas account for 30%, medium-density areas account for 50%, and low-density areas account for 20%.

[0120] To further optimize resource allocation, the distribution ratio is linked to soil nutrient data. Assuming that the average soil nitrogen content in high-density areas is 2.5 mg / kg and that in low-density areas is 1.8 mg / kg, the system automatically generates fertilization recommendations, prioritizing a 10% increase in nitrogen fertilizer input in low-density areas. This forms a complete logical chain from data processing to decision support, ensuring precise planting management.

[0121] Furthermore, if the distribution density of a sub-region in the regionalized distribution ratio information is lower than a preset threshold, the marking mechanism of the low-density region is triggered. The process of generating the marked distribution abnormality data includes:

[0122] By monitoring the distribution density of sub-regions in real time, extracting the density value of each sub-region from the collected data, completing preliminary data sorting, and obtaining the density distribution results of each sub-region;

[0123] If the density distribution result of a sub-area is lower than the preset threshold, the low-density determination process is triggered, and a comparison analysis is performed in combination with historical density data to determine whether the sub-area has a persistent low-density state;

[0124] Based on the low-density status determination results, a marking mechanism is initiated for sub-areas with persistent low density, generating abnormal data records with low-density markers to complete preliminary abnormal marking.

[0125] Adopting the pre-established classification model, the support vector machine algorithm is used to further classify the abnormal data records with low-density markers to determine whether there are potential distribution abnormal patterns;

[0126] Extract relevant features of the distribution anomaly pattern from the classification results, combine them with the regional division information of the sub-region, generate a detailed feature description of the anomaly distribution, and determine the specific range of the anomaly distribution;

[0127] Targeting the specific range of abnormal distribution, relevant environmental variable data are obtained, and correlation analysis between abnormal distribution characteristics and environmental variables is performed through data fusion technology to obtain a comprehensive assessment result of abnormal distribution;

[0128] Based on the comprehensive evaluation results, adjustment strategy data for abnormal distribution is generated. Through the system's automated data update process, the adjustment strategy data is fed back to the density analysis module to complete the dynamic correction of the distribution data.

[0129] For example, by monitoring the distribution density of sub-areas in real time, a sensor network combined with drone image acquisition technology can be used to ensure efficient data acquisition. For example, a 10,000-square-meter area of ​​milk vetch cultivation is divided into 100 sub-areas, each 100 square meters in size. Sensors collect density data every hour, recording the number of plants and converting it into a density value. For example, the density of sub-area B1 is 0.4 plants per square meter. After preliminary data compilation, a density distribution map is generated, clearly showing the density differences between sub-areas.

[0130] It should be noted that the preset threshold can be set to 0.3 plants / square meter. If the density of sub-area B1 falls below this value, the low-density determination process is triggered. Historical density data shows that the average density of B1 over the past three weeks has been 0.35 plants / square meter, confirming the existence of a persistent low-density state. The flagging mechanism is activated and an abnormal data record is generated.

[0131] Specifically, when classifying abnormal data records, the support vector machine algorithm can determine whether there are abnormal distribution patterns based on features such as density, soil moisture, and light intensity. Assuming that soil moisture in B1 is below the normal range, the classification results indicate that this abnormal pattern is related to insufficient water. Characteristic descriptions of abnormal distribution include specific areas, such as B1 and the adjacent sub-areas B2 and B3, covering an area of ​​300 square meters. Fusion analysis of environmental variable data revealed that irrigation water volume in area B1 was only 60% of the normal value, resulting in low density. Comprehensive assessment results suggest increasing irrigation frequency by 20%.

[0132] For example, after generating adjustment strategy data, the system automatically feeds the recommendations back to the density analysis module, updating the irrigation plan. The dynamically corrected density data shows that the density of B1 increased to 0.45 plants / square meter within two weeks. This closed-loop management approach, through real-time monitoring, anomaly classification, and strategic feedback, ensures optimal density distribution across the planting area.

[0133] It should be noted that the correlation analysis of environmental variables can be further combined with historical meteorological data to predict potential anomalies and improve the accuracy of adjustment strategies.

[0134] Optimally, the low-density determination process can incorporate time series analysis to reduce the risk of misjudgment. This multi-level analysis and adjustment mechanism can effectively address distribution anomalies and improve the scientific nature and efficiency of planting management.

[0135] Furthermore, based on the marked distribution abnormality data, image comparison technology is used to match and analyze the plant characteristics of the abnormal area with the pre-established standard image database of milk vetch to determine the growth abnormality and obtain the abnormality cause classification result. The process includes:

[0136] Obtaining a marked abnormal region image from the abnormal distribution data, extracting plant features using an image segmentation algorithm, and obtaining a first feature set;

[0137] Obtaining standard plant features through a preset standard image database of astragalus, and generating a second feature set;

[0138] The cosine similarity algorithm is used to compare the first feature set with the second feature set, calculate the feature matching degree, and determine the feature deviation of the abnormal area;

[0139] If the feature deviation exceeds the preset threshold, the decision tree algorithm is used to classify the deviation data to obtain the preliminary abnormal cause classification results;

[0140] According to the preliminary abnormal cause classification results, the environmental background features are extracted from the abnormal area image to generate an environmental feature set;

[0141] By comparing the environmental feature set with the environmental data in the standard database, the contribution of environmental factors to the abnormal cause is determined, and the final abnormal cause classification result is obtained;

[0142] Based on the final abnormality cause classification results, the Chinese milk vetch standard image database is updated to optimize the accuracy of subsequent feature matching analysis.

[0143] Specifically, this embodiment processes the growth abnormality detection of milk vetch plants. First, real-time image data of milk vetch in the field is obtained through image acquisition equipment. The image resolution is set to 1920x1080 pixels to ensure clear details. Then, the collected image is denoised using an image preprocessing algorithm. The Gaussian filtering method is used with a filter kernel size of 5x5 and a standard deviation of 1.5 to reduce interference from ambient light and debris. Subsequently, based on the abnormal data marking, the abnormal area in the image is extracted, and a threshold-based segmentation algorithm is used to mark pixels with grayscale values ​​below 50 or above 200 as abnormal areas. The proportion of abnormal areas is calculated. For example, if the abnormal pixels in a certain area account for 15% of the total pixels, which exceeds the preset threshold of 10%, it is determined to be abnormal. Next, the extracted plant features of the abnormal area are compared with the pre-established Chinese milk vetch standard image database. The database contains 5,000 standard images, covering normal and various abnormal features (such as yellowing and atrophy). The deep learning model ResNet-50 is used for feature extraction and matching, and the Euclidean distance between the abnormal area and the standard image is calculated. If the distance is less than 0.3, the match is successful and the corresponding abnormality type is output. Finally, the cause of the abnormality is analyzed through the classification results. For example, the matching results show that the yellowing feature accounts for 80%. Combined with the correlation coefficient of 0.85 between yellowing and nitrogen deficiency in the database, it is inferred that the cause of the abnormality is nitrogen deficiency. A classification report is generated, which includes the coordinates of the abnormal area, the type of abnormality and the probability of the cause (such as the probability of nitrogen deficiency is 85%). In order to form a logical chain, if the classification result is uncertain, the soil sensor data can be linked to detect whether the nitrogen content is lower than the standard value of 15mg / kg to further verify the cause and ensure rigorous analysis. The entire process is completed through an automated system, and the data flow and algorithm processing are seamlessly connected to ensure the accuracy of the results.

[0144] Furthermore, the process of quantitative seed reseeding and quantitative application and spraying of special herbicides according to the emergence rate and distribution includes:

[0145] Automatically search for areas that need to be reseeded based on the emergence rate and distribution of seedlings;

[0146] According to the emergence of milk vetch in different areas, quantitative sowing and reseeding are carried out according to the set seed quantity;

[0147] At the same time, based on the distribution information of field plant groups, the usage amount of special herbicides is determined, and the special herbicides are quantitatively used and sprayed according to the set usage amount.

[0148] The present embodiment discloses a method for monitoring and intervening in the cultivation of milk vetch based on image analysis. The method obtains the original image data of the milk vetch planting area through high-resolution image acquisition equipment, and identifies the characteristics of plants at different growth stages through preprocessing and convolutional neural network analysis. Subsequently, the plant distribution density in the sub-area is calculated using a grid division method, abnormal areas are marked, and the causes of abnormal growth are analyzed by comparison with the standard image library. Based on the analysis results, the present invention generates targeted adjustment suggestions and formulates an adaptive intervention strategy in combination with the field management database. Finally, the intervention operation is performed and the monitoring data is dynamically updated. This method realizes intelligent monitoring and precise intervention of the entire process of milk vetch planting, effectively improving the efficiency of planting management and yield quality.

[0149] Example 2

[0150] Based on the same inventive concept, this embodiment further provides a device for identifying and counting the emergence of Chinese milk vetch seedlings, comprising:

[0151] A data storage system for collecting characteristic images of Chinese milk vetch plants at different stages after seedling emergence and establishing an image database;

[0152] Plant information collection and identification system, used to collect images of Chinese milk vetch seedlings after they emerge in the field;

[0153] A data comparison and analysis system is used to compare the collected seedling images with the plant feature images in the image database, calculate the proportion of Chinese milk vetch seedlings in the field plants and the composition of plant groups, and obtain the emergence rate and distribution of Chinese milk vetch;

[0154] The precise material delivery system is used to carry out quantitative seed reseeding and quantitative use and spraying of special herbicides based on the emergence rate and distribution.

[0155] Furthermore, the plant information collection and identification system includes an image collection module and an image processing module;

[0156] The image acquisition module is used to acquire images of the milk vetch seedlings after emergence in the field from multiple angles and different time periods by setting cameras at different positions and angles, thereby obtaining all-round plant characteristic images of the milk vetch seedlings at different stages after emergence in the field;

[0157] The image processing module is used to perform denoising and enhancement processing on the characteristic images of plants at different periods using image preprocessing technology to obtain the morphological characteristics of the milk vetch plants.

[0158] Furthermore, the precise material delivery system includes a quantitative module and a spraying module;

[0159] The quantitative module is used to automatically search for areas that need to be reseeded based on the emergence rate and distribution of the seedlings; quantitative sowing and reseeding are carried out according to the set seed quantity based on the emergence of the milk vetch in different areas;

[0160] The spraying module is used to determine the usage amount of special herbicides according to the distribution information of plant groups in the field, and to quantitatively use and spray the special herbicides according to the set usage amount.

[0161] Furthermore, the device also includes a data update module and a user interaction module;

[0162] The data updating module is used to update and optimize the plant feature image database of the milk vetch at different stages after the emergence of the seedlings based on the actually collected image data;

[0163] The user interaction module is used to display the emergence rate and distribution of Chinese milk vetch, the amount of reseeding seeds, and the amount of herbicide used, and allows users to operate and set.

[0164] The device for identifying and counting the emergence of milk vetch seedlings provided in this embodiment has all the advantages of the method for identifying and counting the emergence of milk vetch seedlings provided in the first embodiment.

[0165] Example 3

[0166] This embodiment further discloses a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method described in the first embodiment.

[0167] Example 4

[0168] This embodiment further discloses a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the method described in the first embodiment are implemented.

[0169] Example 5

[0170] This embodiment further discloses a computer program product, including a computer program, which implements the steps of the method described in the first embodiment when executed by a processor.

[0171] The above are merely preferred embodiments of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A method for identifying and statistically analyzing the emergence of milk vetch seedlings, characterized in that: include: Collect characteristic images of Chinese milk vetch plants at different stages after seedling emergence and establish an image database; Collect images of the milk vetch seedlings after they emerge in the field, compare the collected seedling images with the plant feature images in the image database, calculate the proportion of the milk vetch seedlings in the field plants and the plant group composition, and obtain the emergence rate and distribution of the milk vetch; According to the emergence rate and distribution, quantitative seed reseeding and quantitative use and spraying of special herbicides are carried out.

2. The method according to claim 1, characterized in that The process of collecting characteristic images of Chinese milk vetch plants at different stages after emergence and establishing an image database includes: The cameras were set at different positions and angles to collect images of the milk vetch seedlings in the field from multiple angles and at different time periods after emergence, thus obtaining all-round plant characteristic images of the milk vetch seedlings at different stages after emergence. Image preprocessing technology was used to denoise and enhance the characteristic images of plants at different stages to obtain the morphological characteristics of Chinese milk vetch plants.

3. The method according to claim 2, characterized in that Image preprocessing technology is used to denoise and enhance the characteristic images of plants at different stages. The process of obtaining the morphological characteristics of Chinese milk vetch plants includes: The median filter algorithm is used to denoise the plant feature image to obtain a denoised image dataset; Performing image enhancement processing on the denoised image dataset to obtain an enhanced image dataset; Based on the enhanced image dataset, the K-means clustering algorithm is used to divide the Chinese milk vetch planting area into regions to obtain divided image subsets; Extract plant features from the divided image subsets to obtain a feature vector set; the feature vector set includes leaf morphology, plant height, leaf length-to-width ratio, leaf color depth, edge smoothness, serrations, hairiness, and stem-to-leaf ratio; Classifying the feature vector set by a random forest algorithm to determine the growth stage of the Chinese milk vetch and obtain a growth stage label set; According to the growth stage label set, the changing trends of plant characteristics at different growth stages are analyzed to obtain a characteristic change data set.

4. The method according to claim 3, characterized in that The process of performing image enhancement processing on the denoised image dataset to obtain an enhanced image dataset includes: According to the denoised image dataset, an enhancement operation is used to adjust the contrast and brightness of the image to highlight the morphological characteristics of the milk vetch plant, thereby obtaining a morphological feature image set; For the morphological feature image set, edge detection technology is applied to extract the morphological features of the milk vetch plant and determine the feature-significant area; If the clarity of the feature-salient area is lower than a preset threshold, performing local sharpening processing on the morphological feature image set to obtain a sharpened image set; According to the sharpened image set, the morphological features of the Chinese milk vetch plant are identified using a pre-established classification model to determine the feature category; According to the distribution of feature categories, a cluster analysis method is used to group the sharpened image set to obtain a clear image set; If noise interference exists in the clear image set, a secondary denoising process is performed on the grouped images to determine a final enhanced image data set.

5. The method according to claim 1, wherein The process of calculating the proportion of plants that have emerged from Chinese milk vetch in the field and the composition of plant groups includes: The morphological characteristics of Chinese milk vetch plants were extracted and classified using a convolutional neural network to identify individual plants at different growth stages and obtain classified feature label data. The classified feature label data is combined with a field grid division method to divide the planting area into sub-areas, and the distribution density of the milk vetch plants in each sub-area is calculated to obtain regional distribution ratio information; If the distribution density of a sub-region in the regionalized distribution ratio information is lower than a preset threshold, a marking mechanism for the low-density region is triggered to generate marked distribution abnormality data; According to the marked distribution abnormality data, image comparison technology is used to match and analyze the plant characteristics in the abnormal area with a pre-established standard image database of milk vetch, to determine the growth abnormality and obtain a classification result of the abnormality cause.

6. The method according to claim 1, characterized in that Based on the emergence rate and distribution, the process of quantitative seed reseeding and quantitative application and spraying of special herbicides includes: Automatically search for areas that need to be reseeded based on the emergence rate and distribution of seedlings; According to the emergence of milk vetch in different areas, quantitative sowing and reseeding are carried out according to the set seed quantity; At the same time, based on the distribution information of field plant groups, the usage amount of special herbicides is determined, and the special herbicides are quantitatively used and sprayed according to the set usage amount.

7. A device for identifying and counting the emergence of Chinese milk vetch seedlings, characterized in that: include: A data storage system for collecting characteristic images of Chinese milk vetch plants at different stages after seedling emergence and establishing an image database; Plant information collection and identification system, used to collect images of Chinese milk vetch seedlings after they emerge in the field; A data comparison and analysis system is used to compare the collected seedling images with the plant feature images in the image database, calculate the proportion of Chinese milk vetch seedlings in the field plants and the composition of plant groups, and obtain the emergence rate and distribution of Chinese milk vetch; The material precision delivery system is used to carry out quantitative seed reseeding and quantitative use and spraying of special herbicides according to the emergence rate and distribution.

8. The device according to claim 7, characterized in that The plant information acquisition and identification system includes an image acquisition module and an image processing module; The image acquisition module is used to acquire images of the milk vetch seedlings in the field after emergence from multiple angles and different time periods by setting cameras at different positions and angles, thereby obtaining all-round plant feature images of the milk vetch seedlings at different stages after emergence in the field; The image processing module is used to perform denoising and enhancement processing on plant feature images at different stages using image preprocessing technology to obtain the morphological characteristics of the milk vetch plant.

9. The device according to claim 8, characterized in that The material precision delivery system includes a quantitative module and a spraying module; The quantitative module is used to automatically search for the area that needs to be reseeded according to the emergence rate and distribution of the seedlings; according to the emergence of the milk vetch in different areas, quantitative sowing and reseeding are carried out according to the set seed quantity; The spraying module is used to determine the usage amount of the special herbicide according to the distribution information of the plant groups in the field, and quantitatively use and spray the special herbicide according to the set usage amount.

10. The device according to claim 8, characterized in that The device also includes a data update module and a user interaction module; The data updating module is used to update and optimize the plant feature image database of the Chinese milk vetch at different stages after seedling emergence based on the actually collected image data; The user interaction module is used to display the emergence rate and distribution of Chinese milk vetch, the amount of reseeding seeds, and the amount of herbicide used, and allows the user to operate and set.

Images