Agricultural monitoring system and method based on unmanned aerial vehicle

Abstract

The invention discloses an agricultural monitoring system and method based on an unmanned aerial vehicle, and belongs to the technical field of agricultural intelligent monitoring and artificial intelligence, and the system comprises an unmanned aerial vehicle platform which is equipped with a high-resolution camera and an infrared camera and is used for collecting a multispectral visible light image and a thermal imaging image covering a farmland; a data acquisition module; a data preprocessing module; a super-resolution processing module; the multi-task learning module comprises a shared feature extraction network and a task specific head and is used for detecting and judging the processed image; the feature fusion module is used for fusing the features of the visible light image and the thermal imaging image; a comprehensive analysis and feedback module; and the data storage and review module is used for storing the collected images and analysis results on a cloud or a local server. According to the invention, various farmland conditions can be efficiently and accurately detected, detailed farmland health condition reports and management suggestions are generated, and the efficiency and accuracy of agricultural production are improved.

Description

Technical Field

[0001] The present invention relates to the field of agricultural intelligent monitoring and artificial intelligence technology, and in particular to an agricultural monitoring system and method based on a drone. Background Art

[0002] In recent years, with the advancement of agricultural modernization and the development of drone technology, drones have been used more and more widely in agricultural monitoring. Traditional agricultural monitoring methods mainly rely on manual inspections and fixed ground equipment, which are inefficient, have limited coverage, and are difficult to achieve real-time monitoring and large-scale coverage. With the expansion of farmland scale and the intensification of agricultural production, traditional monitoring methods have been unable to meet the needs of modern agricultural production.

[0003] In existing technologies, drones equipped with high-resolution cameras and infrared cameras for farmland monitoring have gradually become a trend. However, these technologies can usually only achieve single-task monitoring, for example, they can only detect weed coverage or pests and diseases, and cannot achieve multi-task comprehensive monitoring. In addition, existing technologies often face the following problems when processing image data:

[0004] Low data processing efficiency: High-resolution images have a large amount of data, and it is difficult to process them in real time on the drone side, resulting in low data processing efficiency and failure to provide timely feedback on farmland health conditions.

[0005] Insufficient multi-task monitoring capabilities: Existing technologies are usually only capable of monitoring a single task, such as weed detection or pest detection. They lack the ability to comprehensively monitor a variety of farmland conditions and find it difficult to provide a comprehensive assessment of farmland health.

[0006] Lack of infrared data fusion: Many existing systems rely solely on visible light imagery for analysis and fail to fully utilize infrared imagery data, resulting in insufficient detection accuracy in areas such as soil drought and crop nutrient deficiencies.

[0007] Poor real-time performance: Due to the limitations of computing resources and algorithm efficiency, existing systems find it difficult to achieve real-time processing and feedback on the drone side. Farmers and agricultural managers are unable to obtain farmland status information in a timely manner, affecting the timeliness and accuracy of agricultural production decisions. Summary of the invention

[0008] The technical task of the present invention is to address the above shortcomings and provide an agricultural monitoring system and method based on drones, which can efficiently and accurately detect various farmland conditions such as weed coverage, pests and diseases, crop growth, nutrient deficiency and soil drought, and generate detailed farmland health status reports and management recommendations, thereby improving the efficiency and accuracy of agricultural production and reducing labor costs.

[0009] The technical solution adopted by the present invention to solve its technical problem is:

[0010] An agricultural monitoring system based on drones, comprising:

[0011] An unmanned aerial vehicle platform, wherein the unmanned aerial vehicle is equipped with a high-resolution camera and an infrared camera for collecting multi-spectral visible light images and thermal imaging images covering the farmland;

[0012] The data collection module is used to control the drone to fly according to the predetermined route and collect farmland data in real time;

[0013] The data preprocessing module is used to normalize, enhance and spatially align the collected visible light images and thermal imaging images to ensure data consistency and quality;

[0014] Super-resolution processing module, used to improve the resolution and clarity of images, converting low-resolution images into high-resolution images through a convolutional neural network model;

[0015] A multi-task learning module, including a shared feature extraction network and task-specific heads, for detection and judgment of processed images;

[0016] Feature fusion module, used to fuse the features of visible light images and thermal imaging images to improve the accuracy and reliability of analysis;

[0017] Comprehensive analysis and feedback module, used to generate farmland health status report, and transmit it to the ground control station in real time through the communication module, providing instant farmland management suggestions;

[0018] The data storage and review module is used to store the collected images and analysis results in the cloud or local server, and supports the query and analysis functions of historical data to facilitate the review and further analysis of historical data.

[0019] The system integrates super-resolution technology and multi-task learning models to achieve real-time comprehensive monitoring of farmland, and uses infrared data for auxiliary judgment. The system can efficiently and accurately detect various farmland conditions such as weed coverage, pests and diseases, crop growth, nutrient deficiency and soil drought, and generate detailed farmland health status reports and management suggestions, thereby improving the efficiency and accuracy of agricultural production and reducing labor costs.

[0020] Furthermore, the UAV platform includes a computing device and a communication module, the computing device is used to run a super-resolution processing model and a multi-task learning model, and the communication module is used for real-time transmission and reception of data.

[0021] Furthermore, the super-resolution processing module adopts a convolutional neural network model to improve the resolution of the image through multi-layer convolution and upsampling operations;

[0022] The super-resolution processing module runs on the drone side and uses the drone's computing resources to achieve real-time processing.

[0023] Furthermore, the shared feature extraction network in the multi-task learning module is used to extract basic features of the image, and each task-specific head is used to extract information required for a specific task from the shared features;

[0024] The detection and judgment include: weed coverage detection, pest and disease detection, crop growth judgment, nutrient deficiency detection and soil drought detection.

[0025] Furthermore, a shared feature extraction network extracts basic features of the image through a convolutional neural network, which is shared by all tasks; then, an independent task-specific head is designed for each monitoring task (such as weed cover detection, pest and disease detection, crop growth judgment, nutrient deficiency detection, and soil drought detection) to extract the information required for the specific task from the shared feature extraction network.

[0026] Furthermore, the feature fusion module fuses the features of the visible light image and the infrared image through feature splicing and weighted fusion methods to enhance the reliability of the analysis results.

[0027] Furthermore, the comprehensive analysis and feedback module generates a comprehensive farmland health status report based on the analysis results of each task, and sends it to the ground control station in real time through the UAV wireless communication module.

[0028] The present invention also claims a method for agricultural monitoring based on a drone, the implementation of which comprises the following steps:

[0029] 1) Use drones to fly along a predetermined route to collect multispectral visible light images and thermal imaging images covering the farmland;

[0030] 2) Normalize, enhance and spatially align the acquired images;

[0031] 3) Apply super-resolution processing models to improve image resolution;

[0032] 4) Use multi-task learning models to detect weed cover, pests and diseases, crop growth, nutrient deficiency, and soil drought on processed images;

[0033] 5) Fuse the features of visible light images and infrared images to generate comprehensive analysis results;

[0034] 6) Generate a farmland health status report based on the analysis results and transmit it to the ground control station in real time through the communication module to provide immediate farmland management suggestions;

[0035] 7) Store the collected images and analysis results in the cloud or local server to facilitate historical data review and further analysis.

[0036] Furthermore, the super-resolution processing model adopts a convolutional neural network to improve the resolution of the image through multi-layer convolution and upsampling operations.

[0037] Furthermore, the multi-task learning model includes a shared feature extraction network and a task-specific head, which are respectively used to extract basic features of the image and feature information specific to each task.

[0038] Compared with the prior art, the agricultural monitoring system and method based on drone of the present invention has the following beneficial effects:

[0039] First, the present invention greatly improves the efficiency of agricultural monitoring. The drone collects high-resolution visible light and infrared images of farmland in real time, and processes and analyzes them in real time on the drone side, reducing the time and cost of manual inspections. Farmers and agricultural managers can quickly obtain the health status of farmland and make production decisions in a timely manner, thereby improving the efficiency and benefits of agricultural production.

[0040] Secondly, the present invention realizes multi-task comprehensive monitoring. Using a multi-task learning model, the system can simultaneously perform multiple tasks such as weed coverage detection, pest and disease detection, crop growth judgment, nutrient deficiency detection, and soil drought detection, providing a comprehensive assessment of farmland health status. This multi-task comprehensive monitoring capability enables farmers to fully understand all aspects of farmland and ensure the comprehensiveness and scientificity of farmland management.

[0041] In addition, the present invention improves the resolution and clarity of images through super-resolution technology, and improves the accuracy of monitoring results. Combined with the feature fusion of visible light images and infrared images, the system has higher reliability and accuracy in analyzing soil drought and nutrient deficiency. Various problems in farmland can be detected and located more accurately, ensuring that farmland management measures are timely and effective.

[0042] The present invention also implements real-time feedback and decision support functions. The system performs real-time data processing and analysis on the UAV side, and transmits the results to the ground control station in real time through the communication module, providing instant farmland management suggestions, such as irrigation, fertilization, and pest control. This real-time feedback mechanism enables farmers and agricultural managers to respond quickly to various farmland problems and avoid agricultural losses caused by delays.

[0043] At the same time, the present invention reduces equipment and operating costs. By integrating multi-task learning and super-resolution processing technology, efficient computing is achieved on the drone side, reducing the reliance on expensive equipment. The overall cost of the system is reduced, making this efficient agricultural monitoring solution affordable to more farmers.

[0044] Finally, the present invention enhances data storage and analysis capabilities. The collected images and analysis results are stored in the cloud or on a local server to form a system historical database, which is convenient for reviewing and further analyzing historical data. By analyzing historical data, the changing trend of farmland health status can be better evaluated, providing data support for long-term agricultural management. In short, the present invention provides an efficient, accurate, and real-time agricultural monitoring solution by integrating drone technology, super-resolution processing technology, and multi-task learning technology, providing strong technical support for modern agricultural production. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a schematic diagram of a technical solution for establishing an agricultural monitoring system based on a drone provided by an embodiment of the present invention;

[0046] Figure 2 It is a schematic diagram of the technical solution of the operation process of the agricultural monitoring system based on drone provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0047] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

[0048] An agricultural monitoring system based on drones, comprising:

[0049] An unmanned aerial vehicle platform, wherein the unmanned aerial vehicle is equipped with a high-resolution camera and an infrared camera for collecting multi-spectral visible light images and thermal imaging images covering the farmland;

[0050] The data collection module is used to control the drone to fly according to the predetermined route and collect farmland data in real time;

[0051] The data preprocessing module is used to normalize, enhance and spatially align the collected visible light images and thermal imaging images to ensure data consistency and quality;

[0052] Super-resolution processing module, used to improve the resolution and clarity of images, converting low-resolution images into high-resolution images through a convolutional neural network model;

[0053] Multi-task learning module, including shared feature extraction network and task-specific heads, for weed cover detection, pest and disease detection, crop growth judgment, nutrient deficiency detection and soil drought detection in processed images;

[0054] Feature fusion module, used to fuse the features of visible light images and thermal imaging images to improve the accuracy and reliability of analysis;

[0055] Comprehensive analysis and feedback module, used to generate farmland health status report, and transmit it to the ground control station in real time through the communication module, providing instant farmland management suggestions;

[0056] The data storage and review module is used to store the collected images and analysis results in the cloud or local server, and supports the query and analysis functions of historical data to facilitate the review and further analysis of historical data.

[0057] Among them, the drone platform further includes a computing device and a communication module, the computing device is used to run the super-resolution processing model and the multi-task learning model, and the communication module is used for real-time transmission and reception of data.

[0058] The data preprocessing module includes a data normalization unit, a data enhancement unit and a data alignment unit, which are respectively used to perform normalization processing, data enhancement processing and spatial alignment processing on the image data.

[0059] The super-resolution processing module adopts a convolutional neural network model to improve the resolution of the image through multi-layer convolution and upsampling operations.

[0060] The shared feature extraction network in the multi-task learning module is used to extract basic features of the image, and each task-specific head is used to extract information required for a specific task from the shared features.

[0061] The feature fusion module fuses the features of the visible light image and the infrared image through feature splicing and weighted fusion methods to enhance the reliability of the analysis results.

[0062] The comprehensive analysis and feedback module generates a comprehensive farmland health status report based on the analysis results of each task, and sends it to the ground control station in real time through the UAV wireless communication module.

[0063] The data storage and review module supports storing data in a local server or a cloud storage platform, and provides query and analysis functions for historical data.

[0064] The system collects farmland data through drones equipped with high-resolution cameras and infrared cameras, and implements real-time image super-resolution processing and multi-task learning analysis on the drone side to generate farmland health status reports and management recommendations. The system is specifically implemented as follows:

[0065] 1. Data collection:

[0066] First, the drone flies along a predetermined route, equipped with a high-resolution camera and an infrared camera to collect multi-spectral visible light images and thermal imaging images covering the farmland in real time. The high-resolution camera is used to obtain detailed crop growth information, while the infrared camera provides temperature information of the soil and crops. The flight path and data collection frequency of the drone are set according to the size of the farmland and monitoring needs to ensure full coverage and efficient collection. The collected data is transmitted to the computing device in real time through the communication module on the drone for processing.

[0067] 2. Data preprocessing:

[0068] In the data preprocessing stage, the collected visible light and infrared images are normalized to unify the range and format of the image data. Data enhancement increases the diversity of training data and improves the robustness of the model through methods such as rotation, scaling, and cropping. In particular, in order to ensure the spatial alignment of visible light images and infrared images, the two are aligned for subsequent feature fusion and comprehensive analysis. In addition, the data needs to be denoised and improved to ensure the data quality of the input model.

[0069] 3. Super-resolution processing:

[0070] After data preprocessing, the system uses a lightweight super-resolution model to process the image to improve the resolution and clarity of the image; this model is implemented through a convolutional neural network (CNN) to convert low-resolution images into high-resolution images, making subsequent image analysis more accurate. The super-resolution processing module runs on the drone side and uses the drone's computing resources to achieve real-time processing. This step not only improves the image's detail expression, but also improves the accuracy of subsequent analysis tasks.

[0071] 4. Multi-task learning:

[0072] Next, the system uses a multi-task learning model to analyze the processed images. The multi-task learning model consists of a shared feature extraction network and a task-specific head. The shared feature extraction network extracts the basic features of the image through a convolutional neural network, and this part of the network is shared by all tasks. Then, an independent task-specific head is designed for each monitoring task (such as weed cover detection, pest and disease detection, crop growth judgment, nutrient deficiency detection, and soil drought detection) to extract the information required for the specific task from the shared features. Through the feature fusion module, the features of the visible light image and the infrared image are fused to improve the accuracy and reliability of the analysis. The model is trained on the ground and deployed to the computing device of the drone for real-time analysis on the drone side.

[0073] 5. Real-time processing and feedback:

[0074] In actual applications, the computing device on the drone side processes the collected image data in real time and applies super-resolution models and multi-task learning models for analysis. The system generates a report on the health status of farmland through feature fusion and comprehensive judgment. The analysis results are transmitted to the ground control station in real time through the drone's communication module, providing instant farmland management suggestions, such as irrigation, fertilization, and pest control. Real-time processing and feedback functions ensure that farmers and agricultural managers can obtain farmland status information in a timely manner and make accurate production decisions.

[0075] 6. Data storage and review:

[0076] Finally, the system stores the collected images and analysis results in the cloud or on a local server for future review and further analysis. By analyzing historical data, the changing trends of farmland health can be evaluated to provide data support for long-term agricultural management. In addition, the stored data can also be used for continuous improvement and optimization of the model to improve the monitoring accuracy and efficiency of the system.

[0077] Through the above technical solutions, this system realizes an agricultural monitoring system based on drones, which can monitor the weed coverage, pests and diseases, crop growth, nutrient deficiency and soil drought of farmland in real time and efficiently, and provide accurate decision-making support for farmers and managers.

[0078] Combination Figure 2 As shown, the entire operation process of the system is as follows:

[0079] 1. System startup,

[0080] Drone preparation: Make sure the high-resolution camera and infrared camera on the drone are installed and in working order. Check the drone's battery, power supply, and communication module to ensure they are in good condition.

[0081] System initialization: Start the drone’s computing device and load the pre-trained multi-task learning model and super-resolution processing model. After the system initialization is completed, prepare to start data collection and processing.

[0082] 2. Data collection,

[0083] Flight planning: Set the flight path and collection frequency of the drone according to the actual situation of the farmland and monitoring requirements. Ensure that the drone can cover the entire farmland area and collect data at the required time intervals.

[0084] Image acquisition: The drone flies along a predetermined route, using a high-resolution camera to collect multi-spectral visible light images, while using an infrared camera to collect thermal imaging images. All collected data is transmitted to the computing device on the drone in real time.

[0085] 3. Data preprocessing,

[0086] Normalization processing: Normalize the collected visible light and infrared images to unify the range and format of image data and ensure the consistency of input data.

[0087] Data augmentation: Apply data augmentation techniques such as rotation, scaling, and cropping to increase the diversity of the data and improve the robustness and generalization ability of the model.

[0088] Data alignment: Ensure the spatial alignment of visible light images and infrared images to facilitate subsequent feature fusion and comprehensive analysis.

[0089] 4. Super-resolution processing,

[0090] Image processing: Use the super-resolution model to process the pre-processed image to improve the resolution and clarity of the image. This step is performed in real time on the drone side to ensure high data quality.

[0091] Result output: The processed high-resolution images will be used for subsequent multi-task learning analysis to improve the accuracy and precision of the overall analysis.

[0092] 5. Multi-task learning analysis

[0093] Feature extraction: The processed image is subjected to feature extraction through a shared convolutional neural network to obtain multi-level image features, which are shared by all tasks.

[0094] Task-specific analysis: Use an independent task-specific head for each monitoring task (such as weed cover detection, pest and disease detection, crop growth judgment, nutrient deficiency detection, and soil drought detection) to extract the information required for the specific task from the shared features and perform analysis.

[0095] Feature fusion: The features of visible light images and infrared images are fused, and comprehensive judgment is made through the feature fusion module to improve the accuracy and reliability of the analysis.

[0096] 6. Comprehensive analysis and feedback,

[0097] Comprehensive analysis: The analysis results of each task are combined to generate a report on the health status of farmland, including weed coverage, distribution of pests and diseases, crop growth, nutrient deficiency, soil drought and other information.

[0098] Instant feedback: The analysis results are transmitted to the ground control station in real time through the drone’s communication module. The system provides instant farmland management suggestions, such as irrigation, fertilization, and pest control, helping farmers and agricultural managers make quick and accurate decisions.

[0099] 7. Data storage and review,

[0100] Data storage: The collected images and analysis results are stored in the cloud or on a local server to form a historical database of the system for subsequent analysis and decision support.

[0101] Data review: By accessing stored data, users can review historical trends in farmland health status for further analysis and research, providing data support for long-term agricultural management.

[0102] 8. System maintenance and update,

[0103] System maintenance: Regularly check the drone's hardware to ensure the normal operation of the camera, camera, battery, and communication module. Maintain computing equipment to ensure system stability and reliability.

[0104] Model update: Based on new data and technological advances, multi-task learning models and super-resolution models are regularly updated and optimized to improve the system's monitoring accuracy and efficiency.

[0105] This system can solve the problems existing in existing technologies, such as low agricultural monitoring efficiency, limited coverage, poor real-time performance, insufficient multi-task monitoring capabilities, and lack of infrared data fusion. Through this system, the following specific goals can be achieved:

[0106] Improve the efficiency of agricultural monitoring: Use drones to collect high-resolution visible light and infrared images and perform real-time processing on the drone side, which significantly improves the efficiency of agricultural monitoring and reduces the cost and time of manual inspections.

[0107] Realize multi-task comprehensive monitoring: Adopting a multi-task learning model, it can simultaneously perform multiple tasks such as weed cover detection, pest and disease detection, crop growth judgment, nutrient deficiency detection and soil drought detection, providing a comprehensive assessment of farmland health status.

[0108] Improve monitoring accuracy: Super-resolution technology is used to improve image resolution, and combined with the feature fusion of visible light and infrared images, the accuracy and reliability of monitoring results can be improved, especially in terms of soil drought and nutrient deficiency.

[0109] Real-time feedback and decision support: The system can realize real-time data processing and analysis on the UAV side, and transmit the results to the ground control station through the communication module, providing instant farmland management suggestions and supporting farmers and agricultural managers to make fast and accurate production decisions.

[0110] Reduce equipment and operating costs: By integrating multi-task learning and super-resolution processing technology, efficient computing is achieved on the drone side, reducing dependence on expensive equipment and reducing the cost of the overall system.

[0111] Enhanced data storage and analysis capabilities: Collected images and analysis results are stored in the cloud or on local servers to facilitate the review and further analysis of historical data, supporting long-term agricultural management and decision-making optimization.

[0112] By integrating drone technology, super-resolution processing technology and multi-task learning technology, this system provides an efficient, accurate and real-time agricultural monitoring solution, providing strong technical support for modern agricultural production.

[0113] The embodiment of the present invention also provides an agricultural monitoring method based on a drone, and the implementation of the method includes the following steps:

[0114] 1. Use drones to fly along a predetermined route to collect multispectral visible light images and thermal imaging images covering the farmland;

[0115] 2. Normalize, enhance and align the collected images;

[0116] 3. Apply super-resolution processing model to improve the resolution of the image;

[0117] 4. Use multi-task learning models to detect weed cover, pests and diseases, crop growth, nutrient deficiency, and soil drought on processed images;

[0118] 5. Fuse the features of visible light images and infrared images to generate comprehensive analysis results;

[0119] 6. Generate a farmland health status report based on the analysis results and transmit it to the ground control station in real time through the communication module to provide immediate farmland management suggestions;

[0120] 7. Store the collected images and analysis results in the cloud or local server to facilitate historical data review and further analysis.

[0121] Among them, the super-resolution processing model adopts a convolutional neural network to improve the resolution of the image through multi-layer convolution and upsampling operations.

[0122] The multi-task learning model includes a shared feature extraction network and a task-specific head, which are respectively used to extract basic features of the image and feature information specific to each task.

[0123] This method can be implemented based on the drone-based agricultural monitoring system described in the above embodiment.

[0124] First, the drone flies along a predetermined route, equipped with a high-resolution camera and an infrared camera to collect multi-spectral visible light images and thermal imaging images covering the farmland in real time. The high-resolution camera is used to obtain detailed crop growth information, while the infrared camera provides temperature information of the soil and crops. The flight path and data collection frequency of the drone are set according to the size of the farmland and monitoring needs to ensure full coverage and efficient collection. The collected data is transmitted to the computing device in real time through the communication module on the drone for processing.

[0125] In the data preprocessing stage, the collected visible light and infrared images are normalized to unify the range and format of the image data. Data enhancement increases the diversity of training data and improves the robustness of the model through methods such as rotation, scaling, and cropping. In particular, in order to ensure the spatial alignment of visible light images and infrared images, the two are aligned for subsequent feature fusion and comprehensive analysis. In addition, the data needs to be denoised and improved to ensure the data quality of the input model.

[0126] After data preprocessing, the system uses a lightweight super-resolution model to process the image to improve the resolution and clarity of the image; this model is implemented through a convolutional neural network (CNN) to convert low-resolution images into high-resolution images, making subsequent image analysis more accurate. The super-resolution processing module runs on the drone side and uses the drone's computing resources to achieve real-time processing. This step not only improves the image's detail expression, but also improves the accuracy of subsequent analysis tasks.

[0127] Next, the system uses a multi-task learning model to analyze the processed images. The multi-task learning model consists of a shared feature extraction network and a task-specific head. The shared feature extraction network extracts the basic features of the image through a convolutional neural network, and this part of the network is shared by all tasks. Then, an independent task-specific head is designed for each monitoring task (such as weed cover detection, pest and disease detection, crop growth judgment, nutrient deficiency detection, and soil drought detection) to extract the information required for the specific task from the shared features. Through the feature fusion module, the features of the visible light image and the infrared image are fused to improve the accuracy and reliability of the analysis. The model is trained on the ground and deployed to the computing device of the drone for real-time analysis on the drone side.

[0128] In actual applications, the computing device on the drone side processes the collected image data in real time and applies super-resolution models and multi-task learning models for analysis. The system generates a report on the health status of farmland through feature fusion and comprehensive judgment. The analysis results are transmitted to the ground control station in real time through the drone's communication module, providing instant farmland management suggestions, such as irrigation, fertilization, and pest control. Real-time processing and feedback functions ensure that farmers and agricultural managers can obtain farmland status information in a timely manner and make accurate production decisions.

[0129] Finally, the system stores the collected images and analysis results in the cloud or on a local server for future review and further analysis. By analyzing historical data, the changing trends of farmland health can be evaluated to provide data support for long-term agricultural management. In addition, the stored data can also be used for continuous improvement and optimization of the model to improve the monitoring accuracy and efficiency of the system.

[0130] Through the above technical solution, a drone-based agricultural monitoring system has been realized, which can monitor the weed coverage, pests and diseases, crop growth, nutrient deficiency and soil drought of farmland in real time and efficiently, providing accurate decision-making support for farmers and managers.

[0131] This method uses lightweight super-resolution technology to pre-process the collected images to improve the image resolution. Then, a multi-task learning model is used to analyze the images, including weed cover detection, pest and disease detection, crop growth judgment, nutrient deficiency detection, and soil drought detection. The model combines visible light and thermal imaging data to perform feature fusion and comprehensive judgment to improve the accuracy of the analysis. The system realizes real-time processing and feedback on the drone side, generates a detailed farmland health report, and provides management recommendations such as irrigation, fertilization, and pest and disease control. It can effectively improve the efficiency and accuracy of agricultural monitoring and provide reliable decision-making support for farmers and agricultural managers.

[0132] Through the above specific implementations, those skilled in the art can easily implement the present invention. However, it should be understood that the present invention is not limited to the above specific implementations. On the basis of the disclosed implementations, those skilled in the art can arbitrarily combine different technical features to achieve different technical solutions.

[0133] Except for the technical features described in the specification, all other technical features are known technologies to those skilled in the art.

Claims

1. An agricultural monitoring system based on drones, characterized in that: include: UAV platform, the drone is equipped with a high-resolution camera and an infrared camera to collect multi-spectral visible light and thermal imaging images covering the farmland; The data collection module is used to control the drone to fly according to the predetermined route and collect farmland data in real time; The data preprocessing module is used to normalize, enhance and spatially align the collected visible light images and thermal imaging images to ensure data consistency and quality; A super-resolution processing module for converting low-resolution images into high-resolution images through a convolutional neural network model; A multi-task learning module, including a shared feature extraction network and task-specific heads, for detection and judgment of processed images; A feature fusion module is used to fuse the features of the visible light image and the thermal imaging image; Comprehensive analysis and feedback module, used to generate farmland health status report, and transmit it to the ground control station in real time through the communication module, providing instant farmland management suggestions; The data storage and review module is used to store the collected images and analysis results in the cloud or local server, and supports the query and analysis of historical data.

2. The agricultural monitoring system based on drone according to claim 1, characterized in that: The UAV platform includes a computing device and a communication module. The computing device is used to run a super-resolution processing model and a multi-task learning model, and the communication module is used for real-time transmission and reception of data.

3. The agricultural monitoring system based on drone according to claim 1, characterized in that: The super-resolution processing module adopts a convolutional neural network model to improve the resolution of the image through multi-layer convolution and upsampling operations; The super-resolution processing module runs on the drone side and uses the drone's computing resources to achieve real-time processing.

4. The agricultural monitoring system based on drone according to claim 1, characterized in that: The shared feature extraction network in the multi-task learning module is used to extract basic features of the image, and each task-specific head is used to extract information required for a specific task from the shared features; The detection and judgment include: weed coverage detection, pest and disease detection, crop growth judgment, nutrient deficiency detection and soil drought detection.

5. The agricultural monitoring system based on drone according to claim 1 or 4, characterized in that: The shared feature extraction network extracts the basic features of the image through a convolutional neural network, which is shared by all tasks; then, an independent task-specific head is designed for each monitoring task to extract the information required for the specific task from the shared feature extraction network.

6. The agricultural monitoring system based on drone according to claim 1, characterized in that: The feature fusion module fuses the features of the visible light image and the infrared image by means of feature splicing and weighted fusion.

7. The agricultural monitoring system based on drone according to claim 1, characterized in that: The comprehensive analysis and feedback module generates a comprehensive farmland health status report based on the analysis results of each task, and sends it to the ground control station in real time through the UAV wireless communication module.

8. An agricultural monitoring method based on drones, characterized in that: The implementation of this method includes the following steps: 1) Use drones to fly along a predetermined route to collect multispectral visible light images and thermal imaging images covering farmland; 2) Normalize, enhance and spatially align the acquired images; 3) Apply super-resolution processing models to improve image resolution; 4) Use multi-task learning models to detect weed cover, pests and diseases, crop growth, nutrient deficiency, and soil drought on processed images; 5) Fuse the features of visible light images and infrared images to generate comprehensive analysis results; 6) Generate a farmland health status report based on the analysis results and transmit it to the ground control station in real time through the communication module to provide immediate farmland management suggestions; 7) Store the collected images and analysis results in the cloud or local server to facilitate historical data review and further analysis.

9. The agricultural monitoring method based on drone according to claim 8, characterized in that: The super-resolution processing model adopts a convolutional neural network to improve the resolution of the image through multi-layer convolution and upsampling operations.

10. The agricultural monitoring method based on drone according to claim 8, characterized in that: The multi-task learning model includes a shared feature extraction network and a task-specific head, which are respectively used to extract basic features of the image and feature information specific to each task.

Images