Plant factory information automatic acquisition method and system based on unmanned aerial vehicle machine vision

Abstract

The present invention relates to the technical field of agricultural information, and specifically to a plant factory information automatic acquisition method and system based on unmanned aerial vehicle machine vision. The method comprises the following steps: using a camera mounted on an unmanned aerial vehicle to capture an image, and using a deep learning and image segmentation technology to generate a plant region segmentation image. According to the present invention, the unmanned aerial vehicle is used to carry the camera to capture a plant image, and efficient and accurate plant region segmentation is realized by using the deep learning technology. A plant health assessment report is generated by means of a computer vision technology, and a health problem is early warned. A plant health analysis report comprehensively considering environmental factors is generated on the basis of ground sensor data and time series analysis, and comprehensive and accurate health assessment is provided. A multi-dimensional plant health monitoring model provides support for all-round health monitoring. A plant disease nutrition analysis report is generated on the basis of a machine learning classification algorithm, a problem is accurately recognized, and a fertilization and pesticide-spraying strategy is formulated on the basis of a decision support system, thereby optimizing resource utilization, and improving the yield and quality.

Description

Automatic plant factory information collection method and system based on drone machine vision Technical Field

[0001] The present invention relates to the field of agricultural information technology, and in particular to a method and system for automatically collecting plant factory information based on unmanned aerial vehicle (UAV) machine vision. Background Art

[0002] Agricultural information technology is a comprehensive field that uses modern information technology to improve agricultural production efficiency, manage agricultural resources, and optimize decision-making. The development of agricultural information technology aims to promote the modernization and sustainable development of agriculture and provide agricultural producers with more scientific and intelligent tools to adapt to the increasingly complex agricultural environment and demands.

[0003] Among them, the automated plant factory information collection method based on drone machine vision aims to automatically collect information on plant growth and health within the plant factory by combining drone and machine vision technology. Its purpose is to improve plant factory production efficiency and management, providing more accurate and real-time information support for agricultural production. To achieve this goal, the system typically uses drones equipped with various sensors, such as high-resolution cameras and infrared cameras, to monitor and analyze plants within the plant factory using machine learning and computer vision algorithms. This system can obtain real-time information on plant growth status, disease status, nutrient levels, and other information, helping agricultural producers to adjust production strategies and improve crop yield and quality. Compared to traditional manual collection and monitoring methods, it offers advantages such as high efficiency, comprehensiveness, and real-time performance.

[0004] Existing plant factory information collection methods lack depth and breadth in plant health assessment, making it difficult to promptly identify plant changes or potential health issues. Traditional methods often fail to integrate visual and environmental data for comprehensive analysis of environmental factors, resulting in incomplete health assessments. Traditional methods lack effective machine learning support for identifying and classifying plant diseases and nutritional deficiencies, leading to inaccurate disease prevention and nutrient management strategies and inability to fully maximize the effectiveness of pesticides and fertilizers. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the existing technology and propose a method and system for automatically collecting plant factory information based on drone machine vision.

[0006] In order to achieve the above objectives, the present invention adopts the following technical solution: a method for automatically collecting plant factory information based on drone machine vision, comprising the following steps:

[0007] S1: Based on the images captured by the drone-mounted camera, deep learning image segmentation technology is used to generate plant area segmentation images;

[0008] S2: Based on the plant region segmentation image, using computer vision technology to generate a plant health assessment report;

[0009] S3: Based on the plant health assessment report, obtaining data through ground sensors and using time series analysis technology to generate a plant health analysis report;

[0010] S4: Based on the plant health analysis report, a multi-dimensional plant health monitoring model is generated using data fusion technology;

[0011] S5: Based on the multi-dimensional plant health monitoring model, a machine learning classification algorithm is used to generate a plant disease and nutrition analysis report;

[0012] S6: Based on the plant disease and nutrition analysis report, a decision support system is used to generate a pesticide and fertilizer management plan;

[0013] The plant area segmentation image is specifically a division of plant and non-plant areas; the plant health assessment report is specifically an analysis of health indicators such as plant color changes and abnormal leaf texture; the plant health analysis report is specifically a plant health assessment that combines visual data and environmental data; the multi-dimensional plant health monitoring model includes correlation analysis of plant growth conditions, environmental factors, and plant health; the plant disease and nutrition analysis report is specifically an identification and classification of plant diseases and nutritional deficiencies; the pesticide and fertilizer management plan is specifically a fertilization and spraying strategy formulated based on the needs of plants in multiple regions.

[0014] As a further solution of the present invention, based on the image captured by the camera mounted on the drone, the steps of generating the plant area segmentation image using deep learning image segmentation technology are as follows:

[0015] S101: Based on the camera mounted on the drone, it uses optical image stabilization technology to take aerial photos of vegetation and generate aerial images of vegetation;

[0016] S102: Based on the vegetation aerial image, a preprocessed vegetation image is generated by using histogram equalization and color space conversion technology;

[0017] S103: Based on the pre-processed vegetation image, using an image segmentation method based on a convolutional neural network to divide the plant and non-plant areas to generate an initial segmented image;

[0018] S104: Based on the initial segmented image, a plant region segmented image is generated by using morphological filtering and edge smoothing technology.

[0019] As a further solution of the present invention, the steps of generating a plant health assessment report based on plant region segmentation images using computer vision technology are as follows:

[0020] S201: Based on the plant region segmentation image, a plant color analysis report is generated using color analysis and clustering algorithms;

[0021] S202: Based on the plant color analysis report, a plant texture anomaly report is generated using a gray level co-occurrence matrix and texture feature extraction technology;

[0022] S203: Based on the plant texture abnormality report, a support vector machine algorithm is used to evaluate the plant health status and generate a preliminary evaluation report;

[0023] S204: Based on the preliminary assessment report, a plant health assessment report is generated using deep learning and data analysis technology.

[0024] As a further solution of the present invention, based on the plant health assessment report, the steps of generating a plant health analysis report by acquiring data through ground sensors and using time series analysis technology are as follows:

[0025] S301: Based on the plant health assessment report, using data acquisition technology, obtaining environmental data through ground sensors to generate an environmental data set;

[0026] S302: Based on the environmental data set, a time series analysis report is generated by using time series analysis;

[0027] S303: Based on the time series analysis report, generate a plant and environment correlation analysis report by using correlation analysis;

[0028] S304: Based on the plant and environment correlation analysis report, data fusion technology is used to integrate visual and environmental data to generate a plant health assessment report.

[0029] As a further solution of the present invention, based on the plant health analysis report, the steps of generating a multi-dimensional plant health monitoring model using data fusion technology are as follows:

[0030] S401: Based on the plant health assessment report, using data fusion technology to integrate drone and ground sensor data to generate a multi-dimensional plant health monitoring data set;

[0031] S402: Based on the multi-dimensional plant health monitoring data set, a machine learning algorithm is used to perform a correlation analysis between growth conditions and environmental factors, and a plant growth condition and environment correlation analysis report is generated;

[0032] S403: Based on the plant growth status and environment correlation analysis report, a convolutional neural network is used to construct a monitoring model;

[0033] S404: Based on the monitoring model, a multi-dimensional plant health monitoring model is generated using model verification technology.

[0034] As a further solution of the present invention, based on a multi-dimensional plant health monitoring model and using a machine learning classification algorithm, the steps for generating a plant disease nutrition analysis report are as follows:

[0035] S501: Based on the multi-dimensional plant health monitoring model, using data cleaning and standardization, cleaning and standardizing the original plant health data to generate pre-processed plant health data;

[0036] S502: generating plant health characteristic data by principal component analysis based on the preprocessed plant health data;

[0037] S503: Based on the plant health characteristic data, a support vector machine is used to perform classification analysis to generate a preliminary plant disease and nutrition analysis report;

[0038] S504: Based on the preliminary plant disease and nutrition analysis report, a plant disease and nutrition analysis report is generated by using data verification techniques such as a cross-validation method.

[0039] As a further embodiment of the present invention, based on the plant disease and nutrition analysis report, a decision support system is used to generate a pesticide and fertilizer management plan in the following steps:

[0040] S601: Based on the plant disease and nutrition analysis report, data classification technology is used to distinguish multiple types of diseases and nutrition deficiencies, and a detailed plant disease and nutrition requirement report is generated;

[0041] S602: Based on the detailed plant disease nutrient requirement report, a preliminary pesticide and fertilizer application plan is generated using a decision tree or other decision analysis method;

[0042] S603: Based on the preliminary pesticide and fertilizer application plan, a genetic algorithm is used to adapt to the needs of plants in multiple regions and generate an initial pesticide and fertilizer management plan;

[0043] S604: Based on the initial pesticide and fertilizer management plan, a pesticide and fertilizer management plan is generated using simulation evaluation technology.

[0044] An automatic plant factory information collection system based on drone machine vision is used to execute the above-mentioned automatic plant factory information collection method based on drone machine vision. The system includes an image acquisition module, a health assessment module, an environmental data collection module, a data fusion module, a health feature extraction module, a disease and nutrition classification module, and a pesticide and fertilizer application module.

[0045] As a further solution of the present invention, the image acquisition module acquires images based on the camera mounted on the drone, performs histogram equalization and color space conversion, and uses a convolutional neural network to segment plants and non-plant areas to generate a pre-processed vegetation image;

[0046] The health assessment module uses color analysis and clustering algorithms based on pre-processed vegetation images, combined with gray-level co-occurrence matrix and texture feature extraction technology, and uses support vector machines to perform comprehensive plant health assessments and generate plant health assessment reports;

[0047] The environmental data acquisition module obtains environmental data through ground sensors based on the plant health assessment report, performs time series analysis and correlation analysis, and generates a plant and environment correlation analysis report;

[0048] The data fusion module uses a machine learning algorithm to build a monitoring model based on the plant and environment correlation analysis report;

[0049] The health feature extraction module cleans, standardizes and performs principal component analysis on the monitoring data set based on the monitoring model, uses a support vector machine for classification analysis, and generates a plant disease and nutrition analysis report;

[0050] The disease and nutrition classification module uses data classification technology based on plant disease and nutrition analysis reports to distinguish multiple types of diseases and nutrition deficiencies, and uses decision trees to generate preliminary pesticide and fertilizer application plans;

[0051] The pesticide and fertilizer application module uses a genetic algorithm to optimize the management plan based on the preliminary pesticide and fertilizer application plan to generate a pesticide and fertilizer management plan.

[0052] As a further solution of the present invention, the image acquisition module includes an image acquisition submodule, a preprocessing submodule, and a segmentation submodule;

[0053] The health assessment module includes a color analysis submodule, a texture feature extraction submodule, and a support vector machine assessment submodule;

[0054] The environmental data acquisition module includes a data acquisition submodule, a time series analysis submodule, and a correlation analysis submodule;

[0055] The data fusion module includes a data integration submodule, a machine learning modeling submodule, and a data set generation submodule;

[0056] The health feature extraction module includes a data cleaning and standardization submodule, a principal component analysis submodule, and a support vector machine classification submodule;

[0057] The disease nutrition classification module includes a data classification submodule, a decision tree analysis submodule, and an optimization plan submodule;

[0058] The pesticide and fertilizer application module includes a genetic algorithm optimization submodule, a simulation evaluation submodule, and a real-time adjustment submodule.

[0059] Compared with the prior art, the advantages and positive effects of the present invention are:

[0060] In the present invention, images are acquired by a camera mounted on a drone, and deep learning technology is used to segment plant regions, greatly improving the efficiency and accuracy of data collection. Plant regions can be identified quickly and accurately, providing a reliable basis for subsequent health assessments. The plant health assessment report generated using computer vision technology can accurately identify color changes and abnormal leaf textures of plants, and provide early warning of health problems. The plant health analysis report generated by combining ground sensor data and time series analysis technology comprehensively considers environmental factors, making health assessments more comprehensive and accurate. The multi-dimensional plant health monitoring model can comprehensively analyze the correlation between plant growth conditions, environmental factors and plant health, and provide a comprehensive health monitoring for plants. The plant disease and nutrition analysis report generated by the machine learning classification algorithm accurately identifies diseases and nutritional deficiencies, and combined with the pesticide and fertilizer management plan formulated by the decision support system, it can provide targeted fertilization and spraying strategies, optimize resource utilization, and improve the yield and quality of crops. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] FIG1 is a schematic diagram of the workflow of the present invention;

[0062] FIG2 is a flow chart of the refinement of S1 of the present invention;

[0063] FIG3 is a flow chart of the refinement of S2 of the present invention;

[0064] FIG4 is a flow chart of the refinement of S3 of the present invention;

[0065] FIG5 is a flow chart of the refinement of S4 of the present invention;

[0066] FIG6 is a flow chart of the refinement of S5 of the present invention;

[0067] FIG7 is a flow chart of the refinement of S6 of the present invention;

[0068] FIG8 is a system flow chart of the present invention;

[0069] FIG9 is a schematic diagram of the system framework of the present invention. DETAILED DESCRIPTION

[0070] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0071] In the description of the present invention, it should be understood that the terms "length," "width," "up," "down," "front," "back," "left," "right," "vertical," "horizontal," "top," "bottom," "inside," "outside," and the like, indicating positions or relationships, are based on the positions or relationships shown in the accompanying drawings and are intended only to facilitate the description of the present invention and simplify the description. They do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limiting the present invention. Furthermore, in the description of the present invention, "plurality" means two or more, unless otherwise expressly and specifically defined.

[0072] Example

[0073] Referring to FIG1 , the present invention provides a technical solution: a method for automatically collecting plant factory information based on drone machine vision, comprising the following steps:

[0074] S1: Based on the images captured by the drone-mounted camera, deep learning image segmentation technology is used to generate plant area segmentation images;

[0075] S2: Generate a plant health assessment report based on plant region segmentation images using computer vision technology;

[0076] S3: Based on the plant health assessment report, data is obtained through ground sensors and time series analysis technology is used to generate a plant health analysis report;

[0077] S4: Based on the plant health analysis report, data fusion technology is used to generate a multi-dimensional plant health monitoring model;

[0078] S5: Generate plant disease and nutrition analysis reports based on a multi-dimensional plant health monitoring model and a machine learning classification algorithm;

[0079] S6: Generate pesticide and fertilizer management plans based on plant disease and nutrient analysis reports using a decision support system;

[0080] The plant region segmentation image specifically divides the plant and non-plant areas; the plant health assessment report specifically analyzes the health indicators of plant color changes and abnormal leaf texture; the plant health analysis report specifically evaluates plant health by combining visual data and environmental data; the multi-dimensional plant health monitoring model includes correlation analysis of plant growth conditions, environmental factors, and plant health; the plant disease and nutrition analysis report specifically identifies and classifies plant diseases and nutritional deficiencies; the pesticide and fertilizer management plan specifically formulates fertilization and spraying strategies based on the needs of plants in multiple regions.

[0081] Through deep learning image segmentation technology, plants and non-plant areas can be efficiently and accurately identified, providing a clear base image for subsequent health assessments and improving the accuracy and reliability of the data. Computer vision technology is used to conduct health assessments of plant areas, including analysis of health indicators such as plant color changes and leaf texture. Comprehensive assessment reports can fully demonstrate the growth status of plants and provide key information for subsequent environmental data collection. Through ground sensor data acquisition and time series analysis technology, the health status of plants can be understood at a broader level and detailed health analysis reports can be generated to improve the overall understanding of plant health. Through data fusion technology, a multi-dimensional plant health monitoring model is constructed, including correlation analysis of plant growth status, environmental factors and plant health. It can more comprehensively reveal the relationship between plants and the environment, providing a scientific basis for agricultural decision-making; it uses machine learning classification algorithms to generate plant disease and nutrition analysis reports, accurately identifying and classifying plant diseases and nutritional deficiencies; and provides a basis for formulating precise pesticide and fertilizer management plans. Through the application of decision support systems, pesticide and fertilizer management plans formulate fertilization and spraying strategies based on the needs of plants in multiple regions to achieve intelligent agricultural management; it improves the intelligence, precision and sustainability of agricultural production, provides comprehensive data support for agricultural management decisions, and achieves efficient and sustainable agricultural production.

[0082] Please refer to Figure 2. Based on the image captured by the drone-mounted camera, the steps for generating the plant area segmentation image using deep learning image segmentation technology are as follows:

[0083] S101: Based on the camera mounted on the drone, it uses optical image stabilization technology to take aerial photos of vegetation and generate aerial images of vegetation;

[0084] S102: Based on the aerial vegetation image, a pre-processed vegetation image is generated by using histogram equalization and color space conversion technology;

[0085] S103: Based on the pre-processed vegetation image, a convolutional neural network-based image segmentation method is used to divide the plant and non-plant areas to generate an initial segmentation image;

[0086] S104: Based on the initial segmentation image, a plant region segmentation image is generated using morphological filtering and edge smoothing technology.

[0087] Aerial photography of vegetation was performed using a drone equipped with a camera. Optical image stabilization technology ensured that image quality was not affected by flight motion. The drone's flight control system enabled the camera to capture high-resolution images of vegetation in a stable manner. Aerial vegetation images were preprocessed to optimize image quality. Histogram equalization enhanced image contrast, making plant and non-plant areas more distinct. Color space conversion adjusted image color distribution to provide better features for subsequent image segmentation. A deep learning framework, including TensorFlow and PyTorc algorithms, was used in conjunction with convolutional neural network models, including U-Net and SegNe. By training the network on a training dataset, the network was able to learn the characteristics of plant and non-plant areas, perform the segmentation, and generate an initial segmented image. Morphological filtering and edge smoothing techniques were used to optimize the initial segmented image and generate a plant region segmentation image. Morphological filtering removed small noise in the image through erosion and dilation operations, making the plant area clearer. Edge smoothing eliminated the jagged effect of the segmentation edge, making the final generated plant region segmentation image more visually continuous.

[0088] Refer to Figure 3. Based on the plant region segmentation image, the steps for generating a plant health assessment report using computer vision technology are as follows:

[0089] S201: Based on the plant region segmentation image, a plant color analysis report is generated using color analysis and clustering algorithms;

[0090] S202: Based on the plant color analysis report, a plant texture anomaly report is generated using gray-level co-occurrence matrix and texture feature extraction technology;

[0091] S203: Based on the plant texture abnormality report, a support vector machine algorithm is used to evaluate the plant health status and generate a preliminary assessment report;

[0092] S204: Based on the preliminary assessment report, use deep learning and data analysis technology to generate a plant health assessment report.

[0093] Based on the plant region segmentation image, the object region segmentation image is preprocessed, including denoising and brightness adjustment. Color analysis and K-means clustering algorithms are used to obtain plant color distribution information. Color features, including color histograms and average colors, are extracted from the clustering results to generate a plant color analysis report that details the plant color features and their distribution. The plant region image is converted into a grayscale image, and the grayscale co-occurrence matrix is ​​calculated. Then, texture feature extraction technology is used to obtain plant texture information and generate a plant texture anomaly report that describes plant texture anomalies. A support vector machine algorithm is used to classify the labeled plant health data through training. A preliminary assessment report is generated to describe the overall health status of the plant. A convolutional neural network is used to perform data analysis, including feature importance. The output of the deep learning model and the data analysis results are combined to generate the final plant health assessment report.

[0094] Refer to Figure 4. Based on the plant health assessment report, the steps for generating a plant health analysis report using ground sensor data and time series analysis technology are as follows:

[0095] S301: Based on the plant health assessment report, data collection technology is used to obtain environmental data through ground sensors to generate an environmental data set;

[0096] S302: Based on the environmental data set, a time series analysis report is generated using time series analysis;

[0097] S303: Based on the time series analysis report, a correlation analysis report is generated using correlation analysis;

[0098] S304: Based on the plant and environment correlation analysis report, data fusion technology is used to integrate visual and environmental data to generate a plant health assessment report.

[0099] Use ground sensors to acquire environmental data, including key parameters such as soil moisture, temperature, and light. Preprocess the collected data to ensure its accuracy and completeness. Use the collected environmental data to construct a time series model to analyze the trends and periodic changes of each parameter. Use time series analysis techniques to perform trend analysis and anomaly detection to identify abnormalities in the plant growth process. Correlation analysis is performed between key indicators in the plant health assessment report and environmental data to verify the significance of the correlation. Generate a plant-environment correlation analysis report that details the relationship between plant health and environmental factors. Use data fusion technology to integrate environmental data from ground sensors with the plant health assessment report. By comprehensively considering the analysis of visual data and environmental data, a final plant health assessment report is generated, providing detailed information on the correlation between plant health and the environment, including recommendations and improvement measures.

[0100] Refer to Figure 5. Based on the plant health analysis report, the steps for generating a multi-dimensional plant health monitoring model using data fusion technology are as follows:

[0101] S401: Based on the plant health assessment report, data fusion technology is used to integrate drone and ground sensor data to generate a multi-dimensional plant health monitoring dataset;

[0102] S402: Based on the multi-dimensional plant health monitoring data set, a machine learning algorithm is used to perform correlation analysis between growth status and environmental factors, and a plant growth status and environmental correlation analysis report is generated;

[0103] S403: Based on the plant growth status and environment correlation analysis report, a convolutional neural network is used to build a generation monitoring model;

[0104] S404: Based on the monitoring model, a multi-dimensional plant health monitoring model is generated using model verification technology.

[0105] Data fusion technology is used to integrate high-altitude image data acquired by drones with ground sensor data, and the data is preprocessed, including removing outliers and filling missing values, to ensure the accuracy and completeness of the data. Based on the integrated data, a multi-dimensional monitoring data set is constructed, including plant growth indicators and environmental factors. Machine learning algorithms including linear regression and random forest are used to analyze the correlation between growth conditions and environmental factors. Based on the analysis results of the machine learning algorithm, a plant growth condition and environment correlation analysis report is generated. A convolutional neural network model is used to extract features and predict plant health status from multi-dimensional monitoring data. A monitoring data set is constructed, and the CNN model is trained and verified. The model parameters and optimization algorithm are adjusted to obtain the best plant health prediction model. Based on the trained monitoring model, new plant health data are predicted and analyzed; a multi-dimensional plant health monitoring model is generated to provide multi-angle and real-time plant health monitoring.

[0106] Please refer to Figure 6. Based on the multi-dimensional plant health monitoring model and using the machine learning classification algorithm, the steps for generating a plant disease and nutrition analysis report are as follows:

[0107] S501: Based on the multi-dimensional plant health monitoring model, data cleaning and standardization are used to clean and standardize the original plant health data to generate pre-processed plant health data;

[0108] S502: generating plant health characteristic data by principal component analysis based on the preprocessed plant health data;

[0109] S503: Based on the plant health characteristic data, a support vector machine is used to perform classification analysis and generate a preliminary plant disease and nutrition analysis report;

[0110] S504: Based on the preliminary plant disease and nutrition analysis report, a plant disease and nutrition analysis report is generated using data verification techniques such as a cross-validation method.

[0111] A multi-dimensional plant health monitoring model is used to clean and standardize raw plant health data. This includes identifying and processing missing data and outliers, as well as normalizing or standardizing the data to ensure data quality and consistency. Principal component analysis is used to reduce the dimensionality and extract features of the pre-processed plant health data. By identifying the main direction of change in the data, plant health feature data is generated, reducing the data dimension while retaining the main information. Support vector machines are used as a classification algorithm to perform classification analysis using plant health feature data, identifying and distinguishing multiple plant diseases and nutritional status. Data validation techniques such as cross-validation are used to verify the classification results and evaluate the performance and accuracy of the model. Based on the validation results, a plant disease and nutritional analysis report is generated, describing the health of the plant and any existing disease and nutritional issues.

[0112] Refer to Figure 7. Based on the plant disease and nutrition analysis report, the steps for generating a pesticide and fertilizer management plan using a decision support system are as follows:

[0113] S601: Based on the plant disease and nutrition analysis report, data classification technology is used to distinguish multiple types of diseases and nutrition deficiencies, and generate a detailed plant disease and nutrition requirement report;

[0114] S602: Based on the detailed plant disease nutrient requirement report, a preliminary pesticide and fertilizer application plan is generated using decision tree and other decision analysis methods;

[0115] S603: Based on the preliminary pesticide and fertilizer application plan, a genetic algorithm is used to adapt to the needs of plants in multiple regions and generate an initial pesticide and fertilizer management plan;

[0116] S604: Based on the initial pesticide and fertilizer management plan, a pesticide and fertilizer management plan is generated using simulation evaluation technology.

[0117] Use data classification technology to process plant disease nutrition analysis reports, distinguish multiple types of diseases and nutritional deficiencies, use convolutional neural networks or k-nearest neighbor algorithms to generate detailed classification plant disease nutrition requirement reports, and based on the detailed classification plant disease nutrition requirement reports, use decision tree and other decision analysis methods to formulate preliminary pesticide and fertilizer application plans. Use genetic algorithms to optimize the preliminary pesticide and fertilizer application plans to adapt to the plant needs of multiple regions, and use simulation evaluation technology to verify and adjust the final pesticide and fertilizer management plan. By simulating multiple scenarios and environmental conditions, evaluate the effectiveness of the management plan to ensure its applicability under various plant needs and environmental changes.

[0118] Please refer to Figure 8. The automatic plant factory information collection system based on drone machine vision is used to execute the above-mentioned automatic plant factory information collection method based on drone machine vision. The system includes an image acquisition module, a health assessment module, an environmental data collection module, a data fusion module, a health feature extraction module, a disease and nutrition classification module, and a pesticide and fertilizer application module.

[0119] The image acquisition module acquires images based on the camera mounted on the drone, performs histogram equalization and color space conversion, and uses a convolutional neural network to segment plants and non-plant areas to generate pre-processed vegetation images.

[0120] The health assessment module uses color analysis and clustering algorithms based on pre-processed vegetation images, combined with gray-level co-occurrence matrix and texture feature extraction technology, and uses support vector machines to conduct a comprehensive plant health assessment and generate a plant health assessment report.

[0121] The environmental data acquisition module uses ground sensors to obtain environmental data based on the plant health assessment report, conducts time series analysis and correlation analysis, and generates a plant-environment correlation analysis report;

[0122] The data fusion module uses machine learning algorithms to build a monitoring model based on the plant and environment correlation analysis report;

[0123] The health feature extraction module cleans, standardizes, and performs principal component analysis on the monitoring data set based on the monitoring model, uses a support vector machine for classification analysis, and generates a plant disease and nutrition analysis report;

[0124] The disease and nutrient classification module uses data classification technology based on plant disease and nutrient analysis reports to distinguish multiple types of diseases and nutrient deficiencies, and uses decision trees to generate preliminary pesticide and fertilizer application plans;

[0125] The pesticide and fertilizer application module uses genetic algorithms to optimize the management plan based on the preliminary pesticide and fertilizer application plan and generates a pesticide and fertilizer management plan.

[0126] The image acquisition module uses the camera mounted on the drone to collect plant images efficiently and comprehensively, and combines it with histogram equalization and color space conversion to improve image quality and provide clear and accurate basic data for subsequent analysis. The application of convolutional neural networks makes the segmentation of plants and non-plant areas more accurate, generates pre-processed vegetation images, and provides good input for subsequent modules. The health assessment module combines color analysis, clustering algorithms, gray-level co-occurrence matrix and texture feature extraction technology, as well as support vector machines to conduct a comprehensive assessment of plant health. The comprehensive health assessment report provides a comprehensive understanding of the overall health status of the plant, providing a basis for subsequent environmental data collection and disease and nutrition classification. The environmental data acquisition module obtains environmental data through ground sensors, combines time series analysis and correlation analysis, and generates a plant and environment correlation analysis report. This helps to identify the growth differences of plants under different environmental conditions and provides a scientific basis for precision agriculture. The data fusion module uses machine learning algorithms to build a monitoring model based on the plant and environment correlation analysis report, so that the system can continuously learn and optimize, and improve the accuracy of plant status prediction. The health feature extraction module uses support vector machines for classification through cleaning, standardization, and principal component analysis. Classification analysis generates plant disease and nutrition analysis reports; in-depth understanding of the specific conditions of plants provides a basis for subsequent disease classification and pesticide and fertilizer application; the disease and nutrition classification module adopts data classification technology and uses decision trees to generate preliminary pesticide and fertilizer application plans. It can quickly generate preliminary plans based on actual conditions and improve agricultural production efficiency; the pesticide and fertilizer application module uses genetic algorithms for optimization based on preliminary plans to generate more detailed and reasonable pesticide and fertilizer management plans; it can improve the efficiency of agricultural resource utilization, reduce environmental pollution risks, and provide beneficial support for sustainable agricultural development; it can improve the intelligence, precision and sustainability of agricultural production, and provide comprehensive data support for agricultural management decisions.

[0127] Please refer to Figure 9 , the image acquisition module includes an image acquisition submodule, a preprocessing submodule, and a segmentation submodule;

[0128] The health assessment module includes a color analysis submodule, a texture feature extraction submodule, and a support vector machine assessment submodule;

[0129] The environmental data acquisition module includes a data acquisition submodule, a time series analysis submodule, and a correlation analysis submodule;

[0130] The data fusion module includes a data integration submodule, a machine learning modeling submodule, and a data set generation submodule;

[0131] The health feature extraction module includes data cleaning and standardization submodule, principal component analysis submodule, and support vector machine classification submodule;

[0132] The disease and nutrition classification module includes data classification submodule, decision tree analysis submodule, and optimization plan submodule;

[0133] The pesticide and fertilizer application module includes a genetic algorithm optimization submodule, a simulation evaluation submodule, and a real-time adjustment submodule.

[0134] In the image acquisition module, the image acquisition submodule uses a camera to acquire images; the preprocessing submodule performs histogram equalization and color space conversion; and the segmentation submodule uses a semantic segmentation network in deep learning, including U-Net and Mask R-CNN, to segment plants and non-plant areas, generating preprocessed vegetation images.

[0135] In the health assessment module, the color analysis submodule performs color analysis based on preprocessed vegetation images; the texture feature extraction submodule uses gray-level co-occurrence matrix and texture feature extraction technology including Gabor filters to obtain the texture information of the image; and the support vector machine assessment submodule uses a support vector machine to perform a comprehensive assessment of plant health and generate a plant health assessment report.

[0136] In the environmental data acquisition module, the data acquisition submodule acquires environmental data through ground sensors. The time series analysis submodule uses time series analysis techniques, including the ARIMA model, to perform time series analysis on environmental data. The correlation analysis submodule performs correlation analysis and generates a plant-environment correlation analysis report.

[0137] In the data fusion module, the data integration submodule performs data integration based on the plant-environment correlation analysis report, using data integration technologies including data warehouse and ETL; the machine learning modeling submodule uses machine learning algorithms including decision trees, random forests, and support vector machines to build a monitoring model; the dataset generation submodule is responsible for generating the dataset required for the monitoring model.

[0138] In the health feature extraction module, the data cleaning and standardization submodule cleans and standardizes the monitoring data set. The principal component analysis submodule performs principal component analysis to extract the key features of the monitoring data. The support vector machine classification submodule uses a support vector machine to perform classification analysis and generate a plant disease and nutritional analysis report.

[0139] In the disease and nutrient classification module, the data classification submodule uses data classification technology to classify plant disease and nutrient analysis reports. The decision tree analysis submodule uses decision trees to analyze multiple disease types and nutrient deficiencies to generate preliminary pesticide and fertilizer application plans. The optimization plan submodule uses genetic algorithms to optimize the management plan based on the preliminary plan and generate a pesticide and fertilizer management plan.

[0140] In the pesticide and fertilizer application module, the genetic algorithm optimization submodule uses a genetic algorithm to optimize the initial pesticide and fertilizer application plan. The simulation evaluation submodule simulates and evaluates the plan. The real-time adjustment submodule is responsible for adjusting the plan in real time based on actual conditions.

[0141] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. An automatic information collection method for plant factories based on UAV machine vision, characterized in that, it includes the following steps: Based on the images captured by the camera carried by the UAV, using deep learning image segmentation technology to generate a plant area segmentation image; Based on the plant area segmentation image, using computer vision technology to generate a plant health assessment report; Based on the plant health assessment report, obtaining data through ground sensors and using time series analysis technology to generate a plant health analysis report; Based on the plant health analysis report, using data fusion technology to generate a multi-dimensional plant health monitoring model; Based on the multi-dimensional plant health monitoring model, using machine learning classification algorithms to generate a plant disease and nutrition analysis report; Based on the plant disease and nutrition analysis report, using a decision support system to generate a pesticide and fertilizer management plan; The plant area segmentation image is specifically the division of plant and non-plant areas; the plant health assessment report is specifically the analysis of plant health indicators such as color change and abnormal leaf texture; the plant health analysis report is specifically the plant health assessment combining visual data and environmental data; the multi-dimensional plant health monitoring model includes the correlation analysis of plant growth status, environmental factors, and plant health; the plant disease and nutrition analysis report is specifically the identification and classification of plant diseases and nutrient deficiencies; the pesticide and fertilizer management plan is specifically the fertilization and spraying strategy formulated according to the needs of plants in multiple regions.

2. The automatic information collection method for plant factories based on UAV machine vision according to claim 1, characterized in that, The step of generating a plant area segmentation image based on the images captured by the camera carried by the UAV and using deep learning image segmentation technology is specifically: Based on the camera carried by the UAV, using optical image stabilization technology for aerial photography of vegetation to generate a vegetation aerial photography image; Based on the vegetation aerial photography image, using histogram equalization and color space conversion technology to generate a preprocessed vegetation image; Based on the preprocessed vegetation image, using an image segmentation method based on a convolutional neural network to divide plant and non-plant areas and generate an initial segmentation image; Based on the initial segmentation image, using morphological filtering and edge smoothing technology to generate a plant area segmentation image.

3. The automatic information collection method for plant factories based on UAV machine vision according to claim 1, characterized in that, The step of generating a plant health assessment report based on the plant area segmentation image and using computer vision technology is specifically: Based on the plant area segmentation image, using color analysis and clustering algorithms to generate a plant color analysis report; Based on the plant color analysis report, using gray level co-occurrence matrix and texture feature extraction technology to generate a plant texture abnormality report; Based on the plant texture abnormality report, using a support vector machine algorithm to evaluate the plant health status and generate a preliminary assessment report; Based on the preliminary assessment report, using deep learning and data analysis technology to generate a plant health assessment report.

4. The automatic information collection method for plant factories based on UAV machine vision according to claim 1, characterized in that, Based on the plant health assessment report, the steps of obtaining data through ground sensors and generating a plant health analysis report using time series analysis technology are as follows: Based on the plant health assessment report, using data acquisition technology, obtain environmental data through ground sensors to generate an environmental data set; Based on the environmental data set, use time series analysis to generate a time series analysis report; Based on the time series analysis report, use correlation analysis to generate an analysis report on the correlation between plants and the environment; Based on the analysis report on the correlation between plants and the environment, use data fusion technology to integrate visual and environmental data to generate a plant health assessment report.

5. The automatic plant factory information acquisition method based on UAV machine vision according to claim 1, wherein, Based on the plant health analysis report, the steps of generating a multi-dimensional plant health monitoring model using data fusion technology are as follows: Based on the plant health assessment report, use data fusion technology to integrate UAV and ground sensor data to generate a multi-dimensional plant health monitoring data set; Based on the multi-dimensional plant health monitoring data set, use machine learning algorithms to perform correlation analysis on growth conditions and environmental factors to generate an analysis report on the correlation between plant growth conditions and the environment; Based on the analysis report on the correlation between plant growth conditions and the environment, use a convolutional neural network to construct and generate a monitoring model; Based on the monitoring model, use model verification technology to generate a multi-dimensional plant health monitoring model.

6. The automatic plant factory information acquisition method based on UAV machine vision according to claim 1, wherein, Based on the multi-dimensional plant health monitoring model, the steps of generating a plant disease and nutrition analysis report using machine learning classification algorithms are as follows: Based on the multi-dimensional plant health monitoring model, use data cleaning and standardization to clean and standardize the original plant health data to generate preprocessed plant health data; Based on the preprocessed plant health data, use principal component analysis to generate plant health feature data; Based on the plant health feature data, use a support vector machine to perform classification analysis to generate a preliminary analysis report on plant diseases and nutrition; Based on the preliminary analysis report on plant diseases and nutrition, use data verification technologies such as the cross-validation method to generate a plant disease and nutrition analysis report.

7. The automatic plant factory information acquisition method based on UAV machine vision according to claim 1, wherein, Based on the plant disease and nutrition analysis report, the steps of generating a pesticide and fertilizer management plan using a decision support system are as follows: Based on the plant disease and nutrition analysis report, use data classification technology to distinguish multiple types of diseases and nutrient deficiencies to generate a plant disease and nutrition demand report with detailed classification; Based on the plant disease and nutrition demand report with detailed classification, use decision analysis methods such as decision trees to generate a preliminary pesticide and fertilizer application plan; Based on the preliminary pesticide and fertilizer application plan, use a genetic algorithm to adapt to the needs of plants in multiple regions to generate an initial pesticide and fertilizer management plan; Based on the initial pesticide and fertilizer management plan, a simulation evaluation technique is used to generate a pesticide and fertilizer management plan.

8. An automatic plant factory information acquisition system based on unmanned aerial vehicle (UAV) machine vision, characterized in that, According to the automatic plant factory information acquisition method based on UAV machine vision according to any one of claims 1-7, the system includes an image acquisition module, a health assessment module, an environmental data acquisition module, a data fusion module, a health feature extraction module, a disease and nutrition classification module, and a pesticide and fertilizer application module.

9. The automatic plant factory information acquisition system based on UAV machine vision according to claim 8, characterized in that, The image acquisition module acquires images by means of a camera mounted on a UAV, performs histogram equalization and color space conversion, and uses a convolutional neural network to segment plant and non-plant areas to generate a preprocessed vegetation image; The health assessment module is based on the preprocessed vegetation image, uses color analysis and clustering algorithms, combines gray-level co-occurrence matrix and texture feature extraction techniques, and uses a support vector machine to comprehensively evaluate plant health to generate a plant health assessment report; The environmental data acquisition module is based on the plant health assessment report, obtains environmental data through ground sensors, performs time series analysis and correlation analysis, and generates an analysis report on the correlation between plants and the environment; The data fusion module is based on the analysis report on the correlation between plants and the environment, uses machine learning algorithms to construct a monitoring model; The health feature extraction module is based on the monitoring model, cleans, standardizes and performs principal component analysis on the monitoring data set, and uses a support vector machine for classification analysis to generate a plant disease and nutrition analysis report; The disease and nutrition classification module is based on the plant disease and nutrition analysis report, uses data classification techniques to distinguish multiple types of diseases and nutrient deficiencies, and uses a decision tree to generate a preliminary pesticide and fertilizer application plan; The pesticide and fertilizer application module is based on the preliminary pesticide and fertilizer application plan, uses a genetic algorithm to optimize the management plan, and generates a pesticide and fertilizer management plan.

10. The automatic plant factory information acquisition system based on UAV machine vision according to claim 8, characterized in that, The image acquisition module includes an image acquisition sub-module, a preprocessing sub-module, and a segmentation sub-module; The health assessment module includes a color analysis sub-module, a texture feature extraction sub-module, and a support vector machine evaluation sub-module; The environmental data acquisition module includes a data acquisition sub-module, a time series analysis sub-module, and a correlation analysis sub-module; The data fusion module includes a data integration sub-module, a machine learning modeling sub-module, and a data set generation sub-module; The health feature extraction module includes a data cleaning and standardization sub-module, a principal component analysis sub-module, and a support vector machine classification sub-module; The disease and nutrition classification module includes a data classification sub-module, a decision tree analysis sub-module, and an optimization plan sub-module; The pesticide and fertilizer application module includes a genetic algorithm optimization sub-module, a simulation evaluation sub-module, and a real-time adjustment sub-module.

Images