Method, system and storage medium for adjusting working parameters of agricultural machinery

Abstract

The invention relates to the technical field of mechanical parameter adjustment, in particular to a method and system for adjusting operation parameters of agricultural machinery and a storage medium. The method comprises the following steps: acquiring multispectral image data of a farmland through a multispectral camera; performing image segmentation and feature extraction on the multispectral image data, and identifying different soil types, crop growth states and weed distribution to obtain farmland feature map data; performing tillage difficulty prediction of different areas on the farmland feature map data by using a deep learning algorithm to obtain tillage difficulty prediction data; acquiring part image data of the agricultural machinery; and carrying out edge detection and morphological analysis on the image data of the part, and identifying the wear state of the part so as to obtain wear state data. According to the method, deep learning is utilized to carry out tillage difficulty prediction and part wear dynamic modeling, so that the analysis capability of a complex farmland environment and a mechanical state is improved, and the prediction accuracy is enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of mechanical parameter adjustment, and in particular to a method, system and storage medium for adjusting operating parameters of agricultural machinery. Background Art

[0002] Agricultural machinery operating parameters refer to the various technical parameters and operating settings required for agricultural machinery to perform field operations. These parameters usually include the speed of the machine, working depth, working width (this is the forward speed of agricultural machinery when operating in the field. The speed setting is usually adjusted according to the crop type, soil conditions and operation type (such as ploughing, sowing, and harvesting); higher speeds can improve work efficiency, but may affect the quality of work, such as uneven ploughing or inaccurate seed sowing), power demand, fuel consumption (refers to the amount of fuel consumed by agricultural machinery when operating per unit time or per unit area. Fuel consumption directly affects the operating cost and usually depends on the power, speed, workload and engine efficiency of the machine; by optimizing the operating parameters, fuel consumption can be reduced and economic benefits can be improved.), speed, soil contact pressure and walking mode, etc., which have a direct impact on the growth environment, soil conditions and crop yields of crops.

[0003] However, current methods of adjusting the operating parameters of agricultural machinery often have the following problems: soil conditions may vary significantly within the same plot and change over time and weather, and fixed parameter settings are difficult to adapt to all areas, resulting in uneven work quality; for example, the seed drill depth setting may be too deep in soft areas and insufficient in hard areas. As the use time increases, mechanical parts will wear and the original parameter settings may no longer apply; for example, after the plowshare is worn, the actual tillage depth will decrease even if the same depth setting is maintained; or the wear of the rotary tiller blade will affect its cutting performance, requiring constant adjustment of the rotation speed and depth; or the sprinkler nozzles of the sprinkler equipment may become clogged or worn over time, changing the water flow and distribution pattern. Summary of the invention

[0004] Based on this, it is necessary for the present invention to provide a method, system and storage medium for adjusting the operating parameters of agricultural machinery to solve at least one of the above technical problems.

[0005] To achieve the above object, a method for adjusting operating parameters of agricultural machinery comprises the following steps:

[0006] Step S1: acquiring multispectral image data of farmland through a multispectral camera; performing image segmentation and feature extraction on the multispectral image data, and identifying different soil types, crop growth status, and weed distribution to obtain farmland characteristic map data; using a deep learning algorithm to predict the farming difficulty of different areas on the farmland characteristic map data to obtain farming difficulty prediction data;

[0007] Step S2: Acquire component image data of agricultural machinery; perform edge detection and morphological analysis on the component image data, and identify the wear state of the component, thereby obtaining wear state data; establish a component wear dynamics model based on a convolutional neural network according to the wear state data, and perform historical wear trend simulation, thereby obtaining historical wear trend data;

[0008] Step S3: Acquire historical operation log data and current meteorological condition data; perform correlation prediction of operation effect based on historical operation data and current meteorological condition data to obtain operation effect prediction data; perform relationship mapping processing between mechanical operation and component wear on historical operation log data and historical wear trend data to obtain operation-wear mapping data;

[0009] Step S4: Optimize the operation parameters according to the operation-wear mapping data and the operation effect prediction data, and determine the best parameter combination based on the minimum wear, so as to obtain the optimized operation window data; obtain the operation image stream data in real time through the vehicle-mounted camera system; identify the operation depth and crop coverage rate of the operation image stream data, so as to obtain the real-time operation status data;

[0010] Step S5: Generate an optimal parameter adjustment strategy based on the farming difficulty prediction data, the real-time operation status data and the optimized operation window data; adjust the agricultural machinery operation parameters in real time according to the optimal parameter adjustment strategy to achieve dynamic adjustment of the operation parameters.

[0011] The present invention uses multispectral cameras and deep learning technology to more accurately identify different characteristics of farmland, such as soil type, crop growth status and weed distribution. This information is critical for judging the actual condition of farmland and regional differences. Through the prediction of tillage difficulty, the tillage route and operation plan of the machinery can be effectively optimized, thereby reducing unnecessary energy consumption, improving tillage efficiency and reducing the impact on the environment; in addition, this accurate identification and prediction helps to accurately apply fertilizers and control pests and diseases, and reduce production costs. By identifying the wear status of components through edge detection and morphological analysis, the health status of key components of agricultural machinery can be monitored in real time, thereby providing early warning of possible failures. By using convolutional neural networks to establish a wear dynamics model and simulate historical wear trends, future wear conditions can be accurately predicted, the maintenance plan of the machinery can be optimized, unexpected downtime and maintenance costs can be reduced, and equipment utilization and the continuity and stability of operations can be improved. Combining historical operation log data with current meteorological conditions, through correlation analysis and prediction, future operation results can be more accurately evaluated. This correlation prediction can be used to dynamically adjust the operation plan and improve operation efficiency. In addition, by mapping the relationship between operations and component wear, it is possible to identify which operations or conditions have the greatest impact on mechanical wear, thereby optimizing the operation mode of the machinery, reducing wear and extending the life of the equipment. By optimizing the operation parameters and the best parameter combination for minimum wear, the maximum effect of the operation and the minimum wear of the equipment can be achieved, thereby improving the economy and efficiency of mechanical operations. Real-time acquisition of operation image stream data and identification of operation depth and crop coverage can help further verify and adjust the operation parameters to ensure that the operation is carried out according to the optimal strategy. At the same time, this real-time monitoring and adjustment can quickly respond to various changes in the operation process, reduce resource waste and improve output quality. Real-time generation and application of optimal parameter adjustment strategies can dynamically adjust the operation parameters of agricultural machinery according to the real-time input of field conditions and multiple data sources to ensure that the machinery operates in the best state at any time. This dynamic adjustment method improves the flexibility and adaptability of the operation, makes the operation more accurate and efficient, and significantly reduces energy consumption and wear, extends the service life of equipment, and reduces production costs. These steps work together to achieve accurate optimization and dynamic adjustment of agricultural machinery operations using multi-source data, deep learning and machine vision technology. Their effects are mainly reflected in improving operating efficiency, reducing equipment wear, reducing operating costs and environmental impact, and enhancing real-time response and management capabilities for the operating process. These innovative technologies and methods can greatly improve the level of modern agricultural mechanization and promote the development of smart agriculture.

[0012] The present invention also provides a system for adjusting the operating parameters of agricultural machinery, which is used to execute the above method for adjusting the operating parameters of agricultural machinery. The system for adjusting the operating parameters of agricultural machinery includes:

[0013] The farmland analysis module is used to obtain multispectral image data of farmland through a multispectral camera; perform image segmentation and feature extraction on the multispectral image data, and identify different soil types, crop growth status and weed distribution to obtain farmland characteristic map data; use a deep learning algorithm to predict the difficulty of farming in different areas of the farmland characteristic map data to obtain farming difficulty prediction data;

[0014] The mechanical diagnosis module is used to obtain the image data of agricultural machinery components; perform edge detection and morphological analysis on the component image data, and identify the wear state of the components to obtain wear state data; establish a component wear dynamics model based on a convolutional neural network according to the wear state data, and perform historical wear trend simulation to obtain historical wear trend data;

[0015] The operation effect prediction module is used to obtain historical operation log data and current meteorological condition data; perform correlation prediction of operation effects based on historical operation data and current meteorological condition data to obtain operation effect prediction data; perform relationship mapping processing between mechanical operation and component wear on historical operation log data and historical wear trend data to obtain operation-wear mapping data;

[0016] The parameter optimization module is used to optimize the operation parameters according to the operation-wear mapping data and the operation effect prediction data, and determine the best parameter combination based on the minimum wear, so as to obtain the optimized operation window data; obtain the operation image stream data in real time through the vehicle-mounted camera system; identify the operation depth and crop coverage rate of the operation image stream data, so as to obtain the real-time operation status data;

[0017] The dynamic adjustment module is used to generate the optimal parameter adjustment strategy based on the farming difficulty prediction data, the real-time operation status data and the optimized operation window data; the agricultural machinery operation parameters are adjusted in real time according to the optimal parameter adjustment strategy to realize the dynamic adjustment of the operation parameters.

[0018] The farmland analysis module of the present invention obtains multispectral image data of farmland through a multispectral camera, and performs segmentation and feature extraction on it; this process can accurately identify different soil types, crop growth status and weed distribution, and generate farmland feature map data. The use of deep learning algorithms to predict the difficulty of farming in different areas of these data can provide a scientific basis for subsequent operation decisions. On the whole, this module improves the accuracy and speed of farmland condition analysis, provides detailed basic data for the optimization of agricultural machinery operations, and thus improves the level of refined management of agricultural production. The mechanical diagnosis module can quickly identify the wear state of components and generate wear state data by performing edge detection and morphological analysis on the image data of agricultural machinery components. The establishment of a component wear dynamics model based on a convolutional neural network and the simulation of historical wear trends can effectively predict the wear trend of mechanical components under different working conditions. This module can timely detect wear problems of mechanical components, perform preventive maintenance, reduce mechanical failures and unexpected downtime, reduce maintenance costs, extend the service life of equipment, and improve the utilization efficiency and reliability of agricultural machinery. The operation effect prediction module obtains historical operation log data and current meteorological conditions data, and uses machine learning algorithms to predict the association of operation effects, which can provide reliable data-based predictions for future operations. This module also performs relationship mapping processing on historical operation log data and historical wear trend data to generate operation-wear mapping data to help identify the relationship between mechanical operation parameters and component wear. These data can not only be used to optimize mechanical operation parameters and improve operation effects, but also help managers formulate more accurate operation plans and improve the scientific and economic nature of resource allocation. The parameter optimization module builds a multi-objective optimization model based on operation-wear mapping data and operation effect prediction data, and finds the optimal parameter combination through genetic algorithms or other optimization algorithms. This module can minimize mechanical wear and energy consumption while meeting the operation effect requirements, thereby obtaining the optimal operation window data. Real-time acquisition of operation image stream data and depth and crop coverage recognition can ensure the accuracy and real-time nature of operations and improve operation quality and efficiency. Overall, this module can significantly reduce agricultural production costs and improve the comprehensive utilization efficiency of agricultural machinery. The dynamic adjustment module uses the farming difficulty prediction data, real-time operation status data and optimized operation window data to generate the optimal parameter adjustment strategy and adjust the operation parameters of agricultural machinery in real time. Through this dynamic adjustment, the machinery can continuously optimize its operating parameters according to the changes in the actual operating environment to ensure that the operation is always in the optimal state. This module improves the adaptive ability and intelligence level of agricultural machinery, and can maintain stable and efficient operation performance in a changing environment, which helps to further improve the accuracy and quality of agricultural operations.The above five modules form a complete closed-loop control system through the organic combination of data collection, feature analysis, model prediction, parameter optimization and dynamic adjustment, which enables agricultural machinery to automatically adjust and optimize its working parameters under different operating conditions. This not only greatly improves the level of intelligence and automation in agricultural production, but also significantly reduces operating costs, prolongs the life of machinery, improves operating results, and provides strong technical support for the realization of refined management of modern agriculture.

[0019] The present invention also provides a storage medium, comprising:

[0020] Memory for storing computer programs;

[0021] A processor is used to implement the above-mentioned method for adjusting the operating parameters of agricultural machinery when executing the computer program. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments thereof made with reference to the following drawings:

[0023] Figure 1 A schematic diagram of the steps of the method for adjusting the operating parameters of agricultural machinery according to the present invention;

[0024] Figure 2 for Figure 1 Detailed step flow diagram of step S1;

[0025] Figure 3 for Figure 1 Detailed step flow chart of step S2 in FIG. DETAILED DESCRIPTION

[0026] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0027] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0028] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0029] To achieve this, please refer to Figures 1 to 3 The present invention provides a method for adjusting operating parameters of agricultural machinery, the method comprising the following steps:

[0030] Step S1: acquiring multispectral image data of farmland through a multispectral camera; performing image segmentation and feature extraction on the multispectral image data, and identifying different soil types, crop growth status, and weed distribution to obtain farmland characteristic map data; using a deep learning algorithm to predict the farming difficulty of different areas on the farmland characteristic map data to obtain farming difficulty prediction data;

[0031] Step S2: Acquire component image data of agricultural machinery; perform edge detection and morphological analysis on the component image data, and identify the wear state of the component, thereby obtaining wear state data; establish a component wear dynamics model based on a convolutional neural network according to the wear state data, and perform historical wear trend simulation, thereby obtaining historical wear trend data;

[0032] Step S3: Acquire historical operation log data and current meteorological condition data; perform correlation prediction of operation effect based on historical operation data and current meteorological condition data to obtain operation effect prediction data; perform relationship mapping processing between mechanical operation and component wear on historical operation log data and historical wear trend data to obtain operation-wear mapping data;

[0033] Step S4: Optimize the operation parameters according to the operation-wear mapping data and the operation effect prediction data, and determine the best parameter combination based on the minimum wear, so as to obtain the optimized operation window data; obtain the operation image stream data in real time through the vehicle-mounted camera system; identify the operation depth and crop coverage rate of the operation image stream data, so as to obtain the real-time operation status data;

[0034] Step S5: Generate an optimal parameter adjustment strategy based on the farming difficulty prediction data, the real-time operation status data and the optimized operation window data; adjust the agricultural machinery operation parameters in real time according to the optimal parameter adjustment strategy to achieve dynamic adjustment of the operation parameters.

[0035] In the embodiment of the present invention, reference Figure 1 The above is a schematic flow chart of the steps of a method for adjusting the operating parameters of agricultural machinery according to the present invention. In this example, the method for adjusting the operating parameters of agricultural machinery includes the following steps:

[0036] Step S1: acquiring multispectral image data of farmland through a multispectral camera; performing image segmentation and feature extraction on the multispectral image data, and identifying different soil types, crop growth status, and weed distribution to obtain farmland characteristic map data; using a deep learning algorithm to predict the farming difficulty of different areas on the farmland characteristic map data to obtain farming difficulty prediction data;

[0037] The embodiment of the present invention uses a multispectral camera equipped with a high-resolution sensor to comprehensively photograph the farmland and capture image data in the visible, near-infrared and mid-infrared bands. These multispectral image data are input into the image processing software, and image segmentation operations are performed to distinguish the soil types, crop growth status and weed distribution in different regions. Next, feature extraction algorithms such as gray-level co-occurrence matrix (GLCM) and principal component analysis (PCA) are applied to extract key features in the image and form farmland feature map data. Finally, the convolutional neural network (CNN) in a deep learning framework (such as TensorFlow or PyTorch) is used to train and predict these feature map data to obtain the predicted data of the difficulty of farming in each region. These steps quantify the difficulty of farming in different regions through the training and inference process of deep learning, and provide a data basis for subsequent decision-making.

[0038] Step S2: Acquire component image data of agricultural machinery; perform edge detection and morphological analysis on the component image data, and identify the wear state of the component, thereby obtaining wear state data; establish a component wear dynamics model based on a convolutional neural network according to the wear state data, and perform historical wear trend simulation, thereby obtaining historical wear trend data;

[0039] When the embodiment of the present invention obtains the image data of the components of agricultural machinery, it first uses a high-definition camera to capture the image of the components, and ensures that the lighting of the shooting environment is uniform to reduce image noise. These images are input into the image processing software for preprocessing, including noise removal, lighting compensation and geometric correction. Next, an edge detection algorithm, such as Canny edge detection, is applied to extract the component contour information, and morphological analysis techniques, such as opening and closing operations, are used to refine and optimize the edge features to obtain component morphological feature data. Based on these data, a convolutional neural network (CNN) model (such as VGG or ResNet) is used for training to establish a component wear dynamics model, and historical wear trend data is extracted by simulating historical wear data. These steps help to monitor the wear of components in real time and predict their future wear trends.

[0040] Step S3: Acquire historical operation log data and current meteorological condition data; perform correlation prediction of operation effect based on historical operation data and current meteorological condition data to obtain operation effect prediction data; perform relationship mapping processing between mechanical operation and component wear on historical operation log data and historical wear trend data to obtain operation-wear mapping data;

[0041] The embodiment of the present invention extracts historical operation log data from the operation management system of agricultural machinery, including operation time, location, operation type, mechanical parameter settings, operation effect evaluation and meteorological conditions. These data are combined with the current meteorological condition data, and a machine learning model, such as a regression model or a random forest, is used to perform correlation prediction of the operation effect. The correlation between the historical operation data and the meteorological condition data is analyzed using feature engineering technology to obtain operation effect prediction data. At the same time, the historical operation log data is time-aligned with the historical wear trend data, and the time series analysis technology is used to map the data into a time series data set. Finally, the relationship mapping processing of mechanical operation and component wear is performed to obtain operation-wear mapping data. This process provides a deep understanding of the relationship between operation effect and mechanical wear, and provides data support for subsequent optimization.

[0042] Step S4: Optimize the operation parameters according to the operation-wear mapping data and the operation effect prediction data, and determine the best parameter combination based on the minimum wear, so as to obtain the optimized operation window data; obtain the operation image stream data in real time through the vehicle-mounted camera system; identify the operation depth and crop coverage rate of the operation image stream data, so as to obtain the real-time operation status data;

[0043] The embodiment of the present invention constructs a multi-objective optimization model based on the operation-wear mapping data and the operation effect prediction data, with the goal of minimizing component wear and maximizing the operation effect. The model is solved using an optimization algorithm, such as a genetic algorithm or a particle swarm optimization (PSO), to generate a series of optimal parameter combinations. Next, the image stream data during the operation is collected in real time by the on-board camera system, and these data are input into the image processing system for analysis. The image stream data is semantically segmented using a deep learning image segmentation algorithm (such as U-Net or Mask R-CNN) to identify the operation area, non-operation area, and crop area. By calculating the operation depth and crop coverage of the segmentation result data, and then applying time series analysis technology to eliminate instantaneous fluctuations, stable real-time operation status data is obtained. This process ensures accurate optimization of operation parameters and real-time adjustment basis.

[0044] Step S5: Generate an optimal parameter adjustment strategy based on the farming difficulty prediction data, the real-time operation status data and the optimized operation window data; adjust the agricultural machinery operation parameters in real time according to the optimal parameter adjustment strategy to achieve dynamic adjustment of the operation parameters.

[0045] The embodiment of the present invention integrates the farming difficulty prediction data in step S1, the real-time operation status data in step S4, and the optimized operation window data as input features to construct a parameter adjustment strategy generation model based on deep reinforcement learning. The model is trained using a reinforcement learning framework (such as OpenAI Gym), and a reward function is set to minimize wear and maximize the operation effect. After obtaining the current state data, the optimal parameter adjustment strategy is generated through the trained model. Finally, the operating parameters of agricultural machinery are adjusted in real time according to the generated optimal parameter adjustment strategy, and dynamic adjustment is implemented. This process can continuously optimize and adapt to changing operating conditions through a deep reinforcement learning algorithm, thereby realizing dynamic and real-time adjustment of operating parameters and improving the operating efficiency and performance of agricultural machinery.

[0046] The present invention uses multispectral cameras and deep learning technology to more accurately identify different characteristics of farmland, such as soil type, crop growth status and weed distribution. This information is critical for judging the actual condition of farmland and regional differences. Through the prediction of tillage difficulty, the tillage route and operation plan of the machinery can be effectively optimized, thereby reducing unnecessary energy consumption, improving tillage efficiency and reducing the impact on the environment; in addition, this accurate identification and prediction helps to accurately apply fertilizers and control pests and diseases, and reduce production costs. By identifying the wear status of components through edge detection and morphological analysis, the health status of key components of agricultural machinery can be monitored in real time, thereby providing early warning of possible failures. By using convolutional neural networks to establish a wear dynamics model and simulate historical wear trends, future wear conditions can be accurately predicted, the maintenance plan of the machinery can be optimized, unexpected downtime and maintenance costs can be reduced, and equipment utilization and the continuity and stability of operations can be improved. Combining historical operation log data with current meteorological conditions, through correlation analysis and prediction, future operation results can be more accurately evaluated. This correlation prediction can be used to dynamically adjust the operation plan and improve operation efficiency. In addition, by mapping the relationship between operations and component wear, it is possible to identify which operations or conditions have the greatest impact on mechanical wear, thereby optimizing the operation mode of the machinery, reducing wear and extending the life of the equipment. By optimizing the operation parameters and the best parameter combination for minimum wear, the maximum effect of the operation and the minimum wear of the equipment can be achieved, thereby improving the economy and efficiency of mechanical operations. Real-time acquisition of operation image stream data and identification of operation depth and crop coverage can help further verify and adjust the operation parameters to ensure that the operation is carried out according to the optimal strategy. At the same time, this real-time monitoring and adjustment can quickly respond to various changes in the operation process, reduce resource waste and improve output quality. Real-time generation and application of optimal parameter adjustment strategies can dynamically adjust the operation parameters of agricultural machinery according to the real-time input of field conditions and multiple data sources to ensure that the machinery operates in the best state at any time. This dynamic adjustment method improves the flexibility and adaptability of the operation, makes the operation more accurate and efficient, and significantly reduces energy consumption and wear, extends the service life of equipment, and reduces production costs. These steps work together to achieve accurate optimization and dynamic adjustment of agricultural machinery operations using multi-source data, deep learning and machine vision technology. Their effects are mainly reflected in improving operating efficiency, reducing equipment wear, reducing operating costs and environmental impact, and enhancing real-time response and management capabilities for the operating process. These innovative technologies and methods can greatly improve the level of modern agricultural mechanization and promote the development of smart agriculture.

[0047] Preferably, step S1 comprises the following steps:

[0048] Step S11: deploying a multispectral camera array on the target farmland to obtain multi-band image data of the farmland, wherein the multi-band includes a visible light band, a near infrared band, and a mid-infrared band;

[0049] Step S12: Preprocessing the multi-band image data based on denoising, contrast enhancement, and illumination effect correction to obtain multispectral image data;

[0050] Step S13: Calculate the vegetation index and soil index according to the multispectral image data to obtain farmland index map data;

[0051] Step S14: performing image segmentation and feature extraction on the multispectral image data, and identifying different soil types, crop growth status and weed distribution to obtain farmland characteristic map data;

[0052] Step S15: using geographic information system technology to perform spatial analysis and interpolation processing on the farmland characteristic map data, thereby obtaining high-precision farmland characteristic map data;

[0053] Step S16: performing spatial cluster analysis on the high-precision farmland characteristic map data, and dividing farmland areas with similar characteristics, thereby obtaining farmland area division data;

[0054] Step S17: Input the high-precision farmland feature map data into the pre-trained deep learning model, and predict the farming difficulty of the farmland area division data, so as to obtain the farming difficulty prediction data.

[0055] As an embodiment of the present invention, refer to Figure 2 As shown, Figure 1 Detailed step flow diagram of step S1 in the embodiment of the present invention, step S1 includes the following steps:

[0056] Step S11: deploying a multispectral camera array on the target farmland to obtain multi-band image data of the farmland, wherein the multi-band includes a visible light band, a near infrared band, and a mid-infrared band;

[0057] In the embodiment of the present invention, a multispectral camera array is arranged in the target farmland, including multiple cameras, each camera is configured with filters of different bands to capture image data of visible light, near infrared and mid-infrared bands. Ensure that the camera array covers the entire farmland, and calibrate during the shooting process to ensure the geometric consistency and band alignment of the image. Through this array deployment, multi-band image data of each band can be obtained simultaneously, which provides rich farmland information and is helpful for subsequent image analysis and feature extraction.

[0058] Step S12: Preprocessing the multi-band image data based on denoising, contrast enhancement, and illumination effect correction to obtain multispectral image data;

[0059] The embodiment of the present invention imports the acquired multi-band image data into image processing software, such as ENVI or MATLAB. These images are denoised, and a filter (such as a median filter or a Gaussian filter) is used to remove noise in the image. Then contrast enhancement is performed to improve the detail display of the image, using a histogram equalization or adaptive histogram equalization (CLAHE) method. Next, the illumination effect is corrected, and illumination compensation techniques (such as color space conversion and standardization) are used to eliminate the effects of illumination changes. After these processing steps, the corrected multispectral image data is obtained, which improves the image quality and enhances the accuracy of subsequent analysis.

[0060] Step S13: Calculate the vegetation index and soil index according to the multispectral image data to obtain farmland index map data;

[0061] The embodiment of the present invention uses the processed multispectral image data to calculate vegetation index (such as normalized vegetation index NDVI) and soil index (such as soil adjusted vegetation index SAVI) through remote sensing software or custom algorithms. These indices are calculated by the formula, where NDVI = (NIR-R) / (NIR+R), R and NIR are the reflectance of red light and near infrared bands respectively. The calculated index data is used to generate farmland index map data, which reflects the growth of vegetation and the characteristics of soil, and provides a basis for subsequent image segmentation and feature extraction.

[0062] Step S14: performing image segmentation and feature extraction on the multispectral image data, and identifying different soil types, crop growth status and weed distribution to obtain farmland characteristic map data;

[0063] The embodiment of the present invention applies image segmentation algorithms, such as K-means clustering or segmentation methods based on region growing, to multispectral image data. First, image segmentation is performed using preprocessed image data to segment different soil types, crop growth states, and weed distribution areas. Next, feature extraction techniques (such as texture analysis and shape analysis) are used to extract feature information in the image. These feature data help identify and mark different farmland areas, forming detailed farmland feature map data.

[0064] Step S15: using geographic information system technology to perform spatial analysis and interpolation processing on the farmland characteristic map data, thereby obtaining high-precision farmland characteristic map data;

[0065] In the embodiment of the present invention, farmland feature map data is input into geographic information system (GIS) software, such as ArcGIS or QGIS, for spatial analysis; interpolation methods (such as Kriging interpolation or inverse distance weighted interpolation) are used to perform spatial processing on the feature map data to generate high-precision farmland feature map data. These techniques improve the accuracy of feature maps by considering spatial location and density of data points, providing more detailed information for further analysis.

[0066] Step S16: performing spatial cluster analysis on the high-precision farmland characteristic map data, and dividing farmland areas with similar characteristics, thereby obtaining farmland area division data;

[0067] The embodiment of the present invention applies spatial clustering analysis algorithms, such as K-means or DBSCAN, to the generated high-precision farmland feature map data; these algorithms divide the farmland into regions with similar characteristics, and the characteristics of each region are as consistent as possible. The farmland regional division data finally obtained shows the regions with similar soil types, crop growth status and weed distribution in the farmland, thereby providing a regional basis for subsequent prediction of farming difficulty.

[0068] Step S17: Input the high-precision farmland feature map data into the pre-trained deep learning model, and predict the farming difficulty of the farmland area division data, so as to obtain the farming difficulty prediction data.

[0069] In the embodiment of the present invention, high-precision farmland feature map data is input into a pre-trained deep learning model, such as a pre-trained convolutional neural network (CNN) model; farmland area division data is input into the model to predict the difficulty of farming. The model is used to analyze the characteristics of different regions to obtain the predicted data of the difficulty of farming for each region. The predicted data provides a farming difficulty assessment for each region, which helps to make targeted adjustments and optimizations in actual operations.

[0070] The present invention acquires image data of different bands (visible light, near infrared, mid-infrared) through a multi-spectral camera array, and can capture a variety of optical properties of farmland, especially in terms of crop health, soil moisture and weed identification. Image data of different bands provide reflection information of crops and soil under different lighting conditions, helping to more accurately identify the growth status, pests and diseases, and moisture content of plants; the acquisition of such multi-band image data provides a solid foundation for subsequent data analysis and prediction of farming difficulty. Denoising, contrast enhancement and illumination correction of multi-band image data can improve the quality and clarity of the image, making the image data more stable and reliable; this preprocessing step can eliminate influencing factors such as ambient light changes and sensor noise, thereby ensuring the accuracy of the subsequent image analysis process and improving the accuracy of the recognition results. By calculating vegetation indexes (such as NDVI, EVI) and soil indexes, the health status of crops, soil moisture and nutrient status can be quickly evaluated. Vegetation index reflects the growth, photosynthesis efficiency and health of crops, and soil index can reveal the texture and moisture status of soil; these indices provide comprehensive quantitative indicators of farmland conditions, which help to more accurately manage crops and plan cultivation. Image segmentation and feature extraction allow the system to identify and distinguish different soil types, crop growth status and weed distribution in farmland. These data help to understand the actual situation of farmland more accurately and support the formulation of refined farmland management strategies, such as precision fertilization, pest control and weed management, thereby improving crop yields and farmland health. Spatial analysis and interpolation processing of geographic information system (GIS) technology can spatially complete and improve the accuracy of farmland feature map data, making the data more comprehensive and detailed. Through interpolation methods, missing data points can be filled to generate more complete farmland maps, which is very important for large-scale farmland monitoring and management. This high-precision data helps to further improve the analysis accuracy and decision support capabilities of farmland status. Through spatial clustering analysis, the system can automatically divide farmland areas with similar characteristics. This allows differentiated management measures to be taken for different regions, such as implementing corresponding farming strategies and management measures for soil types and crop conditions in different regions; this zoning management helps improve the utilization efficiency of agricultural resources, reduce input and waste, and optimize yield and quality. By inputting high-precision farmland feature map data into a pre-trained deep learning model, the difficulty of farming in different regions can be quickly and accurately predicted; based on this prediction result, agricultural machinery can better plan operation paths and adjust operation parameters to adapt to the farming needs of different regions, reduce mechanical wear and fuel consumption, and improve operation efficiency and results. The effect of these steps is to provide a refined farmland management method based on high-precision multispectral image data and deep learning technology.This method achieves comprehensive monitoring and precise analysis of farmland conditions through multi-band image acquisition, preprocessing, feature extraction, spatial analysis and clustering, combined with a deep learning model, ultimately improving the scientificity and accuracy of operation plans, reducing agricultural production costs, increasing yield and quality, and promoting the intelligent and efficient development of modern agriculture.

[0071] Preferably, step S14 comprises the following steps:

[0072] Step S141: performing multi-scale image segmentation on the farmland index map data to obtain preliminary segmentation result data, wherein the multi-scale image segmentation includes small-scale segmentation for identifying small adjustments, medium-scale segmentation for identifying medium-sized features, and large-scale segmentation for identifying large-scale features;

[0073] The embodiment of the present invention applies multi-scale image segmentation technology to farmland index map data, using image processing software such as MATLAB or OpenCV in the multi-scale segmentation algorithm. First, a small-scale segmentation algorithm (such as SLIC superpixel segmentation) is used to identify small features in the image, such as small weed areas. Next, a medium-scale segmentation method (such as graph-based segmentation) is applied to identify medium-sized soil types or crop areas. Finally, a large-scale segmentation technique (such as hierarchical region growth or edge-based segmentation) is used to identify large-scale features, such as crop coverage of the entire farmland. These segmentation techniques help generate preliminary segmentation result data, laying the foundation for subsequent optimization processing.

[0074] Step S142: optimizing the preliminary segmentation result data based on morphological operators to obtain optimized segmentation graph data;

[0075] The embodiment of the present invention performs optimization processing based on morphological operators on the preliminary segmentation result data, using image processing tools (such as scikit-image library in MATLAB or Python). First, a morphological opening operation is applied to remove small noise points and artifacts, and then a hole filling method is used to fill the holes in the region to improve the continuity of the segmentation result. Finally, a morphological smoothing operation (such as expansion and corrosion) is used to smooth the boundaries and improve the quality of the segmentation map. After these morphological processing steps, the optimized segmentation map data is obtained, and the accuracy and reliability of the preliminary segmentation are improved.

[0076] Step S143: extracting texture, shape and spectral features based on the optimized segmentation map data, thereby obtaining farmland feature data;

[0077] The embodiment of the present invention extracts texture, shape and spectral features based on the optimized segmentation map data, using image analysis software (such as MATLAB's Image Processing Toolbox or Python's OpenCV library). First, the texture features of the image are extracted using a texture analysis method (such as a grayscale co-occurrence matrix). Next, a shape feature extraction algorithm (such as a shape moment or a Hu moment) is used to identify the shape features of the region. Finally, the spectral information of the region is obtained by combining spectral features (such as mean spectral analysis). These extracted feature data constitute detailed farmland feature data, which is helpful for subsequent deep learning model training and feature recognition.

[0078] Step S144: construct a deep convolutional neural network model based on the pre-acquired labeled sample data, and input the farmland feature data into the deep convolutional neural network model to identify the soil type, crop growth status and weed distribution to obtain farmland feature map data.

[0079] The embodiment of the present invention constructs a deep convolutional neural network (CNN) model based on the pre-acquired labeled sample data, using a deep learning framework (such as TensorFlow or PyTorch). First, the farmland characteristic data is used as input, and the data is prepared using preprocessing and enhancement techniques (such as normalization and data amplification). Then, a convolutional neural network architecture is designed, including a convolution layer, a pooling layer, and a fully connected layer, and the model is trained to identify soil type, crop growth status, and weed distribution. The model performance is optimized through cross-validation and hyperparameter tuning. Finally, the farmland characteristic data is input into the trained CNN model to obtain farmland characteristic map data, identify the soil type, crop status, and weed distribution in different areas, and provide detailed information for further analysis and decision-making.

[0080] The multi-scale image segmentation technology of the present invention allows for multi-level analysis of farmland image data. By segmenting images at small, medium and large scales, features of different levels and sizes can be identified. For example, small-scale segmentation can capture fine features of crops such as leaves and weeds, medium-scale segmentation can identify the spacing between crop rows or soil block structures in the field, and large-scale segmentation helps to identify large regional features such as vegetation coverage and soil type. This multi-scale analysis enables the system to process microscopic and macroscopic features simultaneously, improves the accuracy and comprehensiveness of identification, and ultimately provides richer and more detailed data information for farmland management. Optimization techniques based on morphological operators (such as expansion, corrosion, opening and closing operations) can remove noise and impurities, smooth object boundaries, fill holes, and thus improve the quality of preliminary segmentation results. Through this optimization, the edge clarity of useful features in the image can be enhanced, the error in segmentation can be reduced, and the integrity and accuracy of the target features can be improved. This helps to more accurately identify and distinguish different areas and objects in the farmland, laying a more reliable foundation for subsequent feature extraction and classification. The extraction of texture, shape and spectral features using optimized segmentation map data can fully capture the physical and spectral properties of various features in the farmland. Texture features can reflect the surface roughness, density and other properties of crops, shape features can help identify the geometric shape of specific crops or weeds, and spectral features provide reflection information in different spectral ranges, revealing information such as plant health and soil moisture. The combination of these multi-dimensional features can improve the accurate identification of soil type, crop growth status and weed distribution, thereby providing more accurate and detailed data support for intelligent farmland management. By using labeled sample data to train the deep convolutional neural network (CNN) model, the model can learn to recognize and classify various patterns and features in farmland feature data, such as different types of soil, crop health status and weed distribution. Deep convolutional neural networks show extremely high accuracy and robustness when processing complex image data, and can effectively reduce the errors caused by feature complexity and data diversity. Finally, this step realizes automated and efficient farmland feature recognition, improves the speed and accuracy of farmland data analysis, and provides an important basis for refined agricultural management decisions. The effect of these steps is mainly to combine multi-scale image segmentation, morphological optimization, feature extraction and deep learning methods to achieve comprehensive analysis and accurate recognition of farmland image data. Multi-scale image segmentation and morphological optimization make feature extraction more accurate and effective; multi-dimensional feature extraction provides rich data dimensions; the application of deep convolutional neural networks greatly improves the accuracy and automation level of feature recognition; the coordinated application of these technologies can significantly improve the level of intelligent farmland management, help realize precision agriculture, and increase yields and resource utilization efficiency.

[0081] Preferably, step S2 comprises the following steps:

[0082] Step S21: Acquire component image data of agricultural machinery;

[0083] Step S22: performing image preprocessing on the component image data to obtain standardized component image data, wherein the image preprocessing includes image denoising, illumination compensation and geometric correction;

[0084] Step S23: performing multi-level edge detection processing on the standardized component image data and extracting component contour information, thereby obtaining component edge feature data;

[0085] Step S24: extracting component shape features based on morphological analysis of component edge feature data through opening operation, hole filling and edge smoothing, thereby obtaining component morphological feature data;

[0086] Step S25: comparing the component edge feature data and the component morphology feature data with a pre-established standard component model to identify the wear state of the component and obtain wear state data;

[0087] Step S26: Establish a component wear dynamics model based on a convolutional neural network according to the wear state data, and perform a historical wear trend simulation to obtain historical wear trend data.

[0088] As an embodiment of the present invention, refer to Figure 3 As shown, Figure 1 Detailed step flow diagram of step S2 in the embodiment of the present invention, step S2 includes the following steps:

[0089] Step S21: Acquire component image data of agricultural machinery;

[0090] The embodiments of the present invention can be implemented by a high-definition camera or industrial camera installed on the machine. Use the camera to capture the image of the component, ensuring that different angles and positions are covered to fully capture the appearance of the component. When acquiring the image, it is necessary to ensure uniform light to avoid image blur or reflection interference. An image acquisition system (such as FLIR or Basler industrial camera) can be connected to a computer, and the component image data can be recorded and stored through dedicated acquisition software (such as LabVIEW or OpenCV) as a basis for subsequent processing.

[0091] Step S22: performing image preprocessing on the component image data to obtain standardized component image data, wherein the image preprocessing includes image denoising, illumination compensation and geometric correction;

[0092] The embodiment of the present invention performs image preprocessing on the acquired component image data, using image processing software (such as MATLAB or OpenCV library in Python). First, a denoising technique (such as Gaussian filtering) is applied to reduce the noise in the image. Then, illumination compensation (such as histogram equalization) is performed to adjust the brightness and contrast of the image. Finally, geometric correction (such as image registration or deformation correction) is performed to correct the image distortion caused by the shooting angle or distortion, and standardized component image data is obtained to ensure the accuracy of subsequent processing.

[0093] Step S23: performing multi-level edge detection processing on the standardized component image data and extracting component contour information to obtain component edge feature data;

[0094] The embodiment of the present invention performs multi-level edge detection processing on the standardized component image data, using an edge detection algorithm (such as Canny edge detection or Sobel operator), which can be implemented in the OpenCV library of MATLAB or Python. First, multi-level edge detection is applied to the image (such as first applying the Sobel operator to detect the rough edge, and then applying the Canny edge detection to obtain the fine edge), so as to extract the contour information of the component. After processing, the component edge feature data is obtained, which is used for further morphological analysis.

[0095] Step S24: extracting component shape features based on morphological analysis of component edge feature data through opening operation, hole filling and edge smoothing, thereby obtaining component morphological feature data;

[0096] The embodiment of the present invention performs morphological analysis on the edge feature data of the component, using morphological operation tools (such as MATLAB's Morphology Toolbox or Python's scikit-image library). First, an opening operation is applied to remove small noise points, then a hole filling method is used to fill the holes in the component outline, and then an edge smoothing technique (such as dilation and erosion operations) is used to smooth the edges, and finally the morphological feature data of the component is obtained. These operations help extract the accurate shape features of the component for subsequent analysis.

[0097] Step S25: comparing the component edge feature data and the component morphology feature data with a pre-established standard component model to identify the wear state of the component and obtain wear state data;

[0098] The embodiment of the present invention compares the edge feature data and morphological feature data of the component with the pre-established standard component model, and uses an image comparison tool (such as MATLAB's Image Registration Toolbox or Python's OpenCV library template matching). The wear state of the component is identified by a matching algorithm (such as feature point-based matching or image contour-based matching). The identification result is compared with the standard model to obtain wear state data. These data are used to evaluate the wear degree and distribution of the component.

[0099] Step S26: Establish a component wear dynamics model based on a convolutional neural network according to the wear state data, and perform a historical wear trend simulation to obtain historical wear trend data.

[0100] The embodiment of the present invention establishes a component wear dynamics model based on a convolutional neural network (CNN) according to the wear state data, using a deep learning framework (such as TensorFlow or PyTorch). First, the CNN model is designed and trained, and the wear state data is used as input to train the model to identify and predict the dynamic behavior of component wear. The model is trained through historical wear data, and historical wear trend simulation (such as time series prediction) is used to generate simulation data. Finally, component wear trend data is obtained, including information such as wear rate and wear pattern, providing data support for the optimization of operating parameters.

[0101] The present invention can monitor the physical state of the mechanical parts in real time by acquiring the image data of the agricultural machinery parts. This step provides basic data for subsequent image processing and analysis, which helps to timely discover the abnormality or wear of the mechanical parts and avoid downtime and repair costs caused by mechanical failures. Image preprocessing technologies such as denoising, illumination compensation and geometric correction can effectively improve the image quality and make the image data more unified and standardized. Denoising eliminates the noise caused by environmental factors (such as dust and light changes), illumination compensation corrects the brightness and contrast differences caused by changes in lighting conditions, and geometric correction corrects the image distortion caused by changes in camera angle and position. These preprocessing steps ensure the accuracy and reliability of subsequent image analysis. Multi-level edge detection processing can identify and extract edge information in component images, including fine cracks and edge damage. By extracting edge feature data, the shape and contour of the component can be more clearly defined, which helps to find tiny wear or cracks that are difficult to detect with the naked eye; this process improves the sensitivity and accuracy of component status detection and makes early fault detection possible. Morphological analysis (including opening operation, hole filling and edge smoothing) further optimizes the edge feature data, removes noise, fills missing areas, and smoothes irregular edges to extract more accurate and complete component shape features. Through this analysis, the true geometric shape of the component can be better reflected, providing more reliable data support for the judgment of the wear state and improving the recognition accuracy. By comparing the currently acquired component edge and morphological feature data with the pre-established standard component model, the wear condition of the component can be quickly identified. The standard model contains the features of new components or components in an ideal state. By comparison, the location, degree and type of wear (such as cracks, breakage, deformation, etc.) can be identified; this comparison method improves the automation and accuracy of identifying the wear state, and helps to detect and prevent potential mechanical failures at an early stage. The component wear dynamics model built based on convolutional neural network (CNN) can analyze and learn component wear state data and establish a mathematical model of the wear process. By simulating and predicting the historical wear trend of components, the development law and future trend of wear can be understood. In this way, it is helpful to formulate targeted maintenance and replacement strategies, optimize the service life of machinery, reduce maintenance costs and downtime, and improve the overall operation efficiency of agricultural machinery. The effect of these steps is reflected in the full process management of image acquisition, preprocessing, feature extraction, state recognition and model building, which realizes the automatic and intelligent monitoring of the state of agricultural machinery parts. Multiple image processing technology improves the accuracy and early detection capability of wear recognition, and the deep learning model makes the wear trend prediction more reliable, which helps to optimize preventive maintenance and machinery management, and ultimately improves the efficiency and safety of agricultural machinery and reduces operating costs.

[0102] Preferably, step S26 comprises the following steps:

[0103] Step S261: construct a convolutional neural network model, the architecture of which includes an input layer, multiple convolutional layers, a pooling layer, a fully connected layer, and an output layer;

[0104] When constructing a convolutional neural network (CNN) model in an embodiment of the present invention, a deep learning framework (such as TensorFlow or PyTorch) is used. First, the network architecture is defined, including an input layer (receiving image data), multiple convolution layers (extracting features, such as using 32, 64, and 128 convolution kernels), a pooling layer (reducing the dimension of the feature map), a fully connected layer (mapping the convolution features to the output category), and an output layer (predicting the wear state). In the convolution layer, the ReLU activation function is used to increase the nonlinear expression capability, and the maximum pooling (Max Pooling) is used in the pooling layer to reduce the dimension and improve the feature abstraction level. Finally, the softmax or sigmoid function is used to perform classification or regression prediction in the output layer. This model architecture provides the basis for subsequent training and optimization.

[0105] Step S262: training the convolutional neural network model using the wear state data volume, and optimizing the model parameters by back propagation algorithm to obtain an initial component wear dynamics model;

[0106] The embodiment of the present invention uses wear state data to train the CNN model, and uses training data (such as wear images and their corresponding labels) to optimize model parameters through the back propagation algorithm. The training process is implemented in TensorFlow or PyTorch, using a gradient descent optimizer (such as Adam or SGD) to adjust the network weights to minimize the loss function (such as the cross entropy loss function). Through iterative training (such as using mini-batch and multiple rounds), the model parameters are continuously adjusted until the training converges, and the initial component wear dynamics model is obtained, which can predict the wear condition under given wear state data.

[0107] Step S263: using a cross-validation technique to optimize the initial component wear dynamics model by adjusting hyperparameters and increasing or decreasing network layers, thereby obtaining a component wear dynamics model;

[0108] The embodiment of the present invention optimizes the initial component wear dynamics model and uses cross-validation techniques (such as K-fold cross-validation) to evaluate the model performance. By training and validating the model on different data subsets, the hyperparameters (such as learning rate, number of convolution layers, number of filters) and network layer configuration (such as the number of neurons in the fully connected layer) are adjusted. Grid search or random search methods are used to adjust the hyperparameters and improve the model performance. The optimized model will perform better, be able to more accurately predict the wear state, and generate the final component wear dynamics model.

[0109] Step S264: collecting wear status and usage time information under different working conditions, so as to construct a historical wear database;

[0110] The embodiment of the present invention collects wear state data and usage time information under different working conditions to establish a historical wear database. The wear data of agricultural machinery under various operating conditions (such as different loads, speeds, and environmental conditions) can be used to record the wear state and usage time of components under each condition. These data are recorded by sensors and stored in a database system (such as MySQL or MongoDB), providing raw data support for historical wear trend analysis.

[0111] Step S265: Apply the component wear dynamics model to the historical wear database for multiple iterative calculations, simulate the wear progression under different working conditions and time spans, extract wear trend features from the simulation results, and obtain historical wear trend data, where the wear trend features include wear rate, accelerated wear point, and critical wear threshold.

[0112] The embodiment of the present invention applies the optimized component wear dynamics model to the historical wear database and performs multiple iterative calculations to simulate the progress of wear. The simulation is performed in TensorFlow or PyTorch, and the wear prediction is generated by inputting data of different working conditions and time spans into the model. The simulation results are analyzed to extract wear trend features (such as wear rate, accelerated wear point, critical wear threshold), which are calculated by statistical analysis tools (such as MATLAB or Pandas library in Python) to finally obtain historical wear trend data, which will help optimize the operating parameters of agricultural machinery.

[0113] The architecture design of the convolutional neural network (CNN) model of the present invention is crucial for the identification and analysis of the wear state of components. Through the structure including an input layer, multiple convolutional layers, a pooling layer, a fully connected layer and an output layer, the model can effectively process the complex features of image data. The convolutional layer can extract local features of the image, the pooling layer reduces the data dimension and reduces the computational burden, the fully connected layer integrates feature information and makes classification decisions, and the output layer provides prediction results; such an architecture enables the model to have efficient feature extraction and recognition capabilities when processing high-dimensional image data, and can accurately judge the wear state of components. The CNN model is trained by wear state data, and the model can learn and identify the characteristic performance of components under different wear states. The back propagation algorithm is used to continuously adjust the parameters of the model to minimize the prediction error of the model, thereby improving the accuracy and stability of the model. The trained initial component wear dynamics model has basic recognition and prediction capabilities, laying the foundation for further optimization and application. The cross-validation technology can effectively evaluate the performance of the model and avoid overfitting by dividing the data set into a training set and a validation set. By adjusting hyperparameters (such as learning rate, regularization parameters, etc.) and optimizing the network layer structure (such as increasing or decreasing the number of layers), the model can better adapt to different types of wear state data, improve generalization ability and prediction accuracy; the optimized model is more stable and reliable, and can adapt to more diverse working conditions and wear characteristics. The construction of a historical wear database provides rich data support for model training and verification. The database contains the wear state and usage time information of components under various working conditions (such as load, speed, ambient temperature, etc.), which enables the model to learn and verify wear characteristics in a variety of real environment data; this helps the model to more accurately simulate and predict wear trends and improve the practical application value of the model. By applying the optimized model to the historical wear database for multiple iterative calculations, the wear process of components under different working conditions and time spans can be simulated. Wear trend features are extracted from the simulation results, such as wear rate (the speed of wear progression), accelerated wear point (the time point when wear begins to accelerate), and critical wear threshold (the critical point at which a component may fail), providing important early warning information and decision-making basis for mechanical maintenance; this can better formulate maintenance strategies, identify potential failure risks in advance, extend mechanical life, and reduce maintenance and replacement costs. The effect of these steps is mainly reflected in the accurate simulation and trend prediction of component wear process by building and optimizing convolutional neural network models and combining multiple iterative calculations of historical data. By learning and identifying complex wear characteristics, the model can identify potential failure risks in advance and provide intelligent maintenance suggestions, thereby optimizing the operating performance and service life of the machinery. The entire process improves the intelligence level of agricultural machinery maintenance, reduces unexpected downtime and maintenance costs, and provides important technical support for lean management.

[0114] Preferably, step S3 comprises the following steps:

[0115] Step S31: Obtain historical operation log data through the operation management system of the agricultural machinery, including operation time, location, operation type, machine parameter settings, operation effect evaluation and meteorological conditions; obtain current meteorological condition data;

[0116] The embodiment of the present invention obtains historical operation log data through the operation management system of agricultural machinery (such as Fleet Management Software); these log data include operation time, location, operation type, mechanical parameter settings (such as speed, load), operation effect evaluation (such as operation quality score) and meteorological conditions (such as temperature, humidity, precipitation). Current meteorological condition data is obtained in real time through meteorological data providers (such as Weather API). During the data collection process, SQL queries are used to extract historical logs from the database, and API interfaces are used to obtain current meteorological data from meteorological services. All data are summarized and organized through a data management platform (such as Tableau or Power BI) for subsequent analysis.

[0117] Step S32: using a machine learning algorithm to perform feature engineering and model training on historical operation data and its corresponding meteorological conditions, thereby obtaining an operation effect prediction model;

[0118] The embodiment of the present invention uses a machine learning algorithm to perform feature engineering and model training on historical operation data and its corresponding meteorological conditions; in a Python environment, a library such as Scikit-learn or XGBoost is used for feature extraction and model training. Feature engineering includes data cleaning, standardization (such as Min-Max standardization), and feature selection (such as PCA). Then, a regression model (such as linear regression, random forest regression) or a classification model (such as support vector machine, gradient boosting machine) is used to model the operation effect. During the training process, the model performance is evaluated by cross-validation, and the model parameters are optimized using the training set and the validation set to obtain an operation effect prediction model.

[0119] Step S33: inputting the current meteorological condition data into the trained operation effect prediction model to perform operation effect prediction, thereby obtaining operation effect prediction data;

[0120] The embodiment of the present invention inputs the current meteorological condition data into the trained operation effect prediction model to predict the operation effect; in Python, the prediction is performed by calling the trained model (such as a model saved by Joblib or Pickle), taking the current meteorological data as input, and generating the operation effect prediction data. These prediction data include possible results of the operation effect (such as operation quality score, efficiency, etc.), and the output of the model is used for decision support.

[0121] Step S34: performing time alignment and feature matching on the historical operation log data and the historical wear trend data to obtain a time series data set;

[0122] The embodiment of the present invention performs time alignment and feature matching on the historical operation log data and the historical wear trend data; the Pandas library in Python is used for data preprocessing, and the historical operation log data and the historical wear trend data are first sorted by timestamp. Alignment is performed using time series analysis tools (such as Statsmodels or Prophet), and the time periods in the two data sets are matched to generate a time series data set. The feature matching process includes merging the data according to the time range to ensure that the operation log data and the wear data are compared at the same time point to obtain a consistent time series data set.

[0123] Step S35: performing relationship mapping processing between mechanical operation and component wear according to the time series data set to obtain operation-wear mapping data.

[0124] The embodiment of the present invention performs a relationship mapping process between mechanical operation and component wear based on a time series data set. In a Python environment, a data analysis tool (such as a clustering algorithm or regression model in Scikit-learn) is used to analyze the time series data set to identify the relationship between mechanical operation parameters and component wear. Operation-wear mapping data is obtained by constructing a mapping model (such as a linear regression model, association rule analysis). These data indicate the wear of components under different operating conditions and provide a basis for optimizing mechanical operation parameters.

[0125] The present invention collects detailed historical operation log data (including operation time, location, type, mechanical parameters, operation effect evaluation, etc.) and meteorological condition data, providing a comprehensive data basis for subsequent data analysis and model training. Historical data can reveal the performance of agricultural machinery under different operating conditions, operation effect evaluation provides a measurement index, and meteorological condition data (such as temperature, humidity, rainfall, etc.) reveals the impact of environmental factors on operation effect. Combined with these data, the system can more accurately predict the effect of future operations, optimize mechanical operation according to current meteorological conditions, and improve operation efficiency and quality. By performing feature engineering and machine learning model training on historical operation data and its corresponding meteorological conditions, a model capable of predicting operation effect is generated. The feature engineering process extracts key influencing factors (such as mechanical parameters, meteorological conditions, etc.) and removes noise and redundant information, making model training more accurate and efficient. The operation effect prediction model obtained by training can identify the main factors affecting the operation effect according to the historical data pattern, provide targeted predictions, provide data support for actual operation decision-making, optimize mechanical parameter settings, and reduce the impact of uncertainty. The current meteorological condition data is input into the operation effect prediction model so that the model can predict the operation effect according to the current environmental conditions. This prediction data helps operators understand the possible operation performance under current meteorological conditions in advance, adjust the operation plan or machine parameter settings, optimize resource utilization, and ensure the best operation results. This real-time prediction function makes agricultural operations more responsive and flexible, and improves production efficiency and benefits. The historical operation log data and wear trend data are time-aligned and feature-matched to generate a unified time series data set. This step ensures the consistency and comparability of the data, allowing the operation data and wear data to be compared and analyzed on the same time dimension. The time series data set reflects the operation history of agricultural machinery in a specific time period and its corresponding wear conditions, providing a solid foundation for subsequent relationship mapping processing and causal analysis, ensuring the accuracy and relevance of data analysis. Through the time series data set, the relationship between mechanical operation and component wear is analyzed to generate operation-wear mapping data. This mapping data reveals the wear pattern and rate of components under different operation types, parameter settings, and environmental conditions, and can identify high-wear risk operations under specific operation conditions. This can help optimize mechanical operation and adjust operation parameters to reduce unnecessary wear and extend the life of mechanical components, thereby reducing maintenance costs and downtime, and improving the efficiency and reliability of overall mechanical operation. The effectiveness of these steps is mainly reflected in the ability to collect, process, analyze and predict data. Through the acquisition, analysis and model training of historical operation and wear data, the system can accurately predict future operation results and identify key factors that affect operation results and component wear. Using this information, mechanical operation strategies can be dynamically adjusted, operating parameters can be optimized, the life of the machinery can be maximized, maintenance costs can be reduced, and the efficiency and benefits of agricultural production can be improved.

[0126] Preferably, step S4 comprises the following steps:

[0127] Step S41: constructing a multi-objective optimization model with minimizing component wear and maximizing operating effect as objective functions according to the operation-wear mapping data and the operation effect prediction data;

[0128] The embodiment of the present invention constructs a multi-objective optimization model with minimizing component wear and maximizing operating effect as the objective function based on the operation-wear mapping data and the operating effect prediction data. Using the optimization library in Python (such as SciPy or Pyomo), the objective function is first defined, including minimizing component wear and maximizing operating effect. The specific implementation of the objective function can be based on the wear rate and operating effect indicators obtained by historical data analysis. These objective functions and constraints (such as mechanical operation restrictions and operating environment constraints) are input into the optimization model, and solved by setting an appropriate optimization algorithm (such as a genetic algorithm) to obtain a non-dominated solution set of the objective function value, that is, the optimal parameter combination set.

[0129] Step S42: solving the multi-objective optimization model using a genetic algorithm to obtain a series of non-dominated solutions, thereby forming an optimal parameter combination set;

[0130] The embodiment of the present invention uses a genetic algorithm to solve the multi-objective optimization model and obtain a series of non-dominated solutions. Use a genetic algorithm library (such as DEAP or PyGMO), set appropriate genetic algorithm parameters (such as population size, crossover probability, mutation probability), and then input the defined multi-objective optimization model into the algorithm for iterative solution. The genetic algorithm will generate a series of non-dominated solutions, which represent different optimal parameter combinations that balance component wear and operating performance. These non-dominated solutions are displayed through visualization tools (such as Matplotlib) to help analyze and select the most suitable parameter combination.

[0131] Step S43: performing cluster analysis on the optimal parameter combination set to determine the optimal parameter combination interval, thereby generating optimized operation window data;

[0132] The embodiment of the present invention performs cluster analysis on the optimal parameter combination set to determine the optimal parameter combination interval, thereby generating the optimized operation window data. The optimal parameter combination set is clustered using a data analysis library in Python (such as K-means or DBSCAN algorithm in Scikit-learn). The interval of the parameter combination is determined based on the clustering results, and the optimized operation window data is generated. This data will be used to adjust the parameters in actual operations to ensure that the mechanical operation can achieve the best performance under various conditions.

[0133] Step S44: collecting image stream data during the operation in real time through the vehicle-mounted camera system;

[0134] The embodiment of the present invention collects image stream data in real time during the operation process through a vehicle-mounted camera system. The vehicle-mounted camera system (such as a high-resolution camera and a real-time video acquisition system) is installed on the agricultural machinery to continuously capture the image stream during the operation process. The image acquisition software (such as OpenCV) is used to process and store the real-time data stream. The image data is transmitted to the central processing unit for further processing and analysis to ensure the high quality and real-time performance of the image data.

[0135] Step S45: using a deep learning image segmentation algorithm to perform semantic segmentation on the image stream data based on the identification of the working area, the non-working area and the crop area, thereby obtaining semantic segmentation result data;

[0136] The embodiment of the present invention uses a deep learning image segmentation algorithm to perform semantic segmentation on the image stream data based on the identification of the working area, the non-working area and the crop area. The real-time image stream is processed using a pre-trained segmentation model (such as U-Net or DeepLab) in a deep learning framework such as TensorFlow or PyTorch. The model is trained and adjusted to identify different area categories (working area, non-working area, crop area) and generate semantic segmentation result data. These data provide detailed information about the working area and crop coverage, providing a basis for subsequent analysis.

[0137] Step S46: Calculate the operation depth and crop coverage rate based on the semantic segmentation result data, perform time series analysis on the calculation results, and eliminate instantaneous fluctuations to obtain real-time operation status data, wherein the operation depth and crop coverage rate are calculated by comparing the operation area in the image with the ground reference object, and the crop coverage rate is calculated by the ratio of the crop area to the total area.

[0138] The embodiment of the present invention calculates the operation depth and crop coverage rate based on the semantic segmentation result data, performs time series analysis on the calculation results, and eliminates instantaneous fluctuations; the operation depth is calculated by comparing the operation area with the ground reference object (such as the calibration plate) through image analysis software (such as OpenCV), and the crop coverage rate is calculated using the ratio of the crop area to the total area. The calculation results are analyzed using time series analysis tools (such as the time series function in Pandas or the time series analysis in Statsmodels), the instantaneous fluctuations in the data are eliminated, and stable real-time operation status data is generated. These data provide in-depth insights into the operation process and help adjust the operation parameters in real time.

[0139] The present invention constructs a multi-objective optimization model, which simultaneously considers two objective functions, namely, minimizing component wear and maximizing operating effect, so that the model can find the best balance between operating efficiency and mechanical durability. This model can integrate operation-wear mapping data and operation effect prediction data, comprehensively analyze the influence of different operating parameters on mechanical performance and life, and help decision makers select the optimal operation plan, while maximizing output efficiency and reducing maintenance costs and equipment loss. Genetic algorithm is a heuristic optimization algorithm, which is particularly suitable for solving multi-objective optimization problems. By solving the multi-objective optimization model by genetic algorithm, a large-scale parameter space can be efficiently explored to obtain a series of non-dominated solutions (i.e., Pareto optimal solutions), which represent parameter selections that achieve optimal balance under different weight combinations. The non-dominated solution set forms a diversified optimal parameter combination set, which provides multiple options for different operating conditions and objectives, and improves the flexibility and accuracy of decision-making. By clustering the optimal parameter combination set, parameter combination intervals with similar characteristics can be identified, thereby generating the optimal operation window data. This step can further simplify the decision-making process, transform complex parameter selection problems into simple interval selections, and improve the intuitiveness and convenience of operation. At the same time, the generated optimized operation window data can be used as a guide for actual operation to ensure that the machinery operates within the most suitable parameter range, achieve the best operation effect and minimize component wear. Real-time collection of image stream data during the operation process provides rich visual information support for subsequent operation status monitoring and analysis. The on-board camera system can capture changes in the operation area in real time, monitor the operation of the machinery and the status of the crop. This real-time data collection method ensures the timeliness and accuracy of the operation status information, and provides reliable basic data for dynamic adjustment and optimization. The deep learning image segmentation algorithm can automatically identify different areas in the image, such as the operation area, the non-operation area, and the crop area, to obtain semantic segmentation result data. This image-based automatic recognition capability greatly improves the efficiency and accuracy of data processing and reduces human judgment errors. The semantic segmentation results can reflect the progress and effect of the operation process in real time, helping operators to understand the operation coverage and quality in real time and make corresponding adjustments. The operation depth and crop coverage are calculated through the semantic segmentation result data, which can accurately quantify various indicators of the operation process; time series analysis helps to eliminate instantaneous fluctuations and provide smoother and more continuous operation status data. These real-time operation status data provide immediate feedback for machine operation, helping operators to adjust operation parameters in time to adapt to the current operating environment and conditions, ensuring that the machine always operates in the best condition, thereby improving operation efficiency and crop yield. The effect of these steps is concentrated on the optimization adjustment and intelligent control of agricultural machinery operation parameters through multi-objective optimization models, intelligent algorithms and real-time data collection and analysis.The entire process, from data collection and modeling, optimization and solution, real-time monitoring to dynamic adjustment, forms a closed-loop feedback system, enabling agricultural machinery to find the best balance between operating efficiency and component life, maximize production benefits, and reduce maintenance costs and resource consumption.

[0140] Preferably, step S5 comprises the following steps:

[0141] Step S51: using the farming difficulty prediction data, the real-time operation status data, and the optimized operation window data as input features to construct a parameter adjustment strategy generation model based on deep reinforcement learning;

[0142] The embodiment of the present invention uses the farming difficulty prediction data, real-time operation status data and optimized operation window data as input features to build a parameter adjustment strategy generation model based on deep reinforcement learning. Use a deep learning framework such as TensorFlow or PyTorch to create a deep reinforcement learning model, and select a suitable model architecture, such as a deep Q network (DQN) or a deep deterministic policy gradient (DDPG). The farming difficulty prediction data, real-time operation status data and optimized operation window data are used as input features of the model, and the model is trained to generate the optimal parameter adjustment strategy by designing a suitable reward function (for example, maximizing the operation effect and minimizing component wear). These feature data can be standardized and processed by feature engineering tools (such as Scikit-learn) in the data preprocessing module.

[0143] Step S52: using historical operation data to train the parameter adjustment strategy generation model, thereby obtaining initialization model parameters;

[0144] The embodiment of the present invention uses historical operation data to train the parameter adjustment strategy generation model to obtain the initialization model parameters. Use historical operation data (including operation time, location, mechanical parameter settings, etc.) as training data, and train these data with the constructed deep reinforcement learning model. Use the training tools in the deep learning framework (such as TensorFlow or PyTorch), use historical data as input, optimize the model parameters through the back propagation algorithm, and obtain the initialization model parameters. During the training process, set reasonable hyperparameters (such as learning rate, batch size) and tune them through the validation set to ensure that the model can learn effective parameter adjustment strategies.

[0145] Step S53: obtaining current state data; using the initialized model parameters and the parameter adjustment strategy generation model to generate an optimal parameter adjustment strategy according to the current state data;

[0146] The embodiment of the present invention obtains current operation status data (such as operation depth, crop coverage, etc.) and real-time environmental data (such as meteorological conditions). These data are input into the initialized deep reinforcement learning model, and the optimal parameter adjustment strategy is generated through forward propagation calculation. The model is inferred in real time using the reasoning tool in the reinforcement learning framework (such as TensorFlow Serving or PyTorch Serve) to obtain the optimized operation parameter adjustment strategy. These strategies will be used to dynamically adjust the operation parameters of agricultural machinery.

[0147] Step S54: adjusting the operating parameters of the agricultural machinery in real time according to the optimal parameter adjustment strategy to achieve dynamic adjustment of the operating parameters.

[0148] The embodiment of the present invention applies the generated optimal parameter adjustment strategy to the control system of agricultural machinery. The operating parameters of the machinery (such as engine speed, operating depth, etc.) are adjusted in real time through the control interface of the agricultural machinery (such as CAN bus interface). The control system can be integrated with the deep learning model through embedded programming (such as using C++ or Python to write control logic) to achieve dynamic adjustment. The feedback from the monitoring system (such as operating results, component status) will be used as the basis for further adjustment to ensure the continuous optimization of operating parameters in actual operation.

[0149] The present invention integrates the farming difficulty prediction data, real-time operation status data and optimized operation window data as input features to construct a parameter adjustment strategy generation model based on deep reinforcement learning. Deep reinforcement learning can adaptively learn the optimal strategy in a complex and dynamic environment, and the model can continuously learn and improve from environmental feedback. This adaptive strategy generation mechanism can enable agricultural machinery to achieve optimal performance under various operating conditions and improve the autonomy and intelligence level of mechanical operation. By training the model with historical operation data, a large amount of known operation information and experience can be effectively utilized to help the model quickly converge to a reasonable initial state. Doing so can reduce the exploration time and learning cost of the model in practical applications and enhance the adaptability and practicality of the model. The effective setting of the initial parameters can improve the reaction speed and decision-making accuracy of the model in actual operations, thereby improving the efficiency and reliability of mechanical operations. Obtaining the current state data enables the model to perceive the actual situation of the operation site in real time and make decisions based on these data. By using the initialization model parameters and the parameter adjustment strategy generation model, the optimal parameter adjustment strategy under the current operating environment can be quickly generated to achieve the real-time response capability of agricultural machinery. This optimization and adjustment strategy based on the current state helps to continuously optimize mechanical performance in a dynamic and uncertain operating environment, ensuring efficient and safe operation. Based on the optimal parameter adjustment strategy, the operating parameters of agricultural machinery are adjusted in real time, so that the machinery can respond quickly and adaptively according to the changing operating conditions on site. This dynamic adjustment mechanism ensures that the machinery always works under the optimal operating parameters, which can maximize the operating effect (such as tillage quality and efficiency) and minimize equipment wear and energy consumption. In addition, the real-time adjustment capability can help cope with emergencies and environmental changes, and improve the flexibility and stability of operations. The above steps realize the adaptive optimization and dynamic adjustment of agricultural machinery operating parameters by combining deep reinforcement learning and real-time data processing. The entire process uses the reinforcement learning model to continuously learn and improve the parameter adjustment strategy, so as to achieve the best performance in a changing agricultural operating environment. This method not only improves the intelligence and automation of agricultural machinery, but also responds quickly under uncertain conditions, improves operating efficiency, reduces maintenance costs, and ensures the continuity and reliability of operations.

[0150] The present invention also provides a system for adjusting the operating parameters of agricultural machinery, which is used to execute the above method for adjusting the operating parameters of agricultural machinery. The system for adjusting the operating parameters of agricultural machinery includes:

[0151] The farmland analysis module is used to obtain multispectral image data of farmland through a multispectral camera; perform image segmentation and feature extraction on the multispectral image data, and identify different soil types, crop growth status and weed distribution to obtain farmland characteristic map data; use a deep learning algorithm to predict the difficulty of farming in different areas of the farmland characteristic map data to obtain farming difficulty prediction data;

[0152] The mechanical diagnosis module is used to obtain the image data of agricultural machinery components; perform edge detection and morphological analysis on the component image data, and identify the wear state of the components to obtain wear state data; establish a component wear dynamics model based on a convolutional neural network according to the wear state data, and perform historical wear trend simulation to obtain historical wear trend data;

[0153] The operation effect prediction module is used to obtain historical operation log data and current meteorological condition data; perform correlation prediction of operation effects based on historical operation data and current meteorological condition data to obtain operation effect prediction data; perform relationship mapping processing between mechanical operation and component wear on historical operation log data and historical wear trend data to obtain operation-wear mapping data;

[0154] The parameter optimization module is used to optimize the operation parameters according to the operation-wear mapping data and the operation effect prediction data, and determine the best parameter combination based on the minimum wear, so as to obtain the optimized operation window data; obtain the operation image stream data in real time through the vehicle-mounted camera system; identify the operation depth and crop coverage rate of the operation image stream data, so as to obtain the real-time operation status data;

[0155] The dynamic adjustment module is used to generate the optimal parameter adjustment strategy based on the farming difficulty prediction data, the real-time operation status data and the optimized operation window data; the agricultural machinery operation parameters are adjusted in real time according to the optimal parameter adjustment strategy to realize the dynamic adjustment of the operation parameters.

[0156] The farmland analysis module of the present invention obtains multispectral image data of farmland through a multispectral camera, and performs segmentation and feature extraction on it; this process can accurately identify different soil types, crop growth status and weed distribution, and generate farmland feature map data. The use of deep learning algorithms to predict the difficulty of farming in different areas of these data can provide a scientific basis for subsequent operation decisions. On the whole, this module improves the accuracy and speed of farmland condition analysis, provides detailed basic data for the optimization of agricultural machinery operations, and thus improves the level of refined management of agricultural production. The mechanical diagnosis module can quickly identify the wear state of components and generate wear state data by performing edge detection and morphological analysis on the image data of agricultural machinery components. The establishment of a component wear dynamics model based on a convolutional neural network and the simulation of historical wear trends can effectively predict the wear trend of mechanical components under different working conditions. This module can timely detect wear problems of mechanical components, perform preventive maintenance, reduce mechanical failures and unexpected downtime, reduce maintenance costs, extend the service life of equipment, and improve the utilization efficiency and reliability of agricultural machinery. The operation effect prediction module obtains historical operation log data and current meteorological conditions data, and uses machine learning algorithms to predict the association of operation effects, which can provide reliable data-based predictions for future operations. This module also performs relationship mapping processing on historical operation log data and historical wear trend data to generate operation-wear mapping data to help identify the relationship between mechanical operation parameters and component wear. These data can not only be used to optimize mechanical operation parameters and improve operation effects, but also help managers formulate more accurate operation plans and improve the scientific and economic nature of resource allocation. The parameter optimization module builds a multi-objective optimization model based on operation-wear mapping data and operation effect prediction data, and finds the optimal parameter combination through genetic algorithms or other optimization algorithms. This module can minimize mechanical wear and energy consumption while meeting the operation effect requirements, thereby obtaining the optimal operation window data. Real-time acquisition of operation image stream data and depth and crop coverage recognition can ensure the accuracy and real-time nature of operations and improve operation quality and efficiency. Overall, this module can significantly reduce agricultural production costs and improve the comprehensive utilization efficiency of agricultural machinery. The dynamic adjustment module uses the farming difficulty prediction data, real-time operation status data and optimized operation window data to generate the optimal parameter adjustment strategy and adjust the operation parameters of agricultural machinery in real time. Through this dynamic adjustment, the machinery can continuously optimize its operating parameters according to the changes in the actual operating environment to ensure that the operation is always in the optimal state. This module improves the adaptive ability and intelligence level of agricultural machinery, and can maintain stable and efficient operation performance in a changing environment, which helps to further improve the accuracy and quality of agricultural operations.The above five modules form a complete closed-loop control system through the organic combination of data collection, feature analysis, model prediction, parameter optimization and dynamic adjustment, which enables agricultural machinery to automatically adjust and optimize its working parameters under different operating conditions. This not only greatly improves the level of intelligence and automation in agricultural production, but also significantly reduces operating costs, prolongs the life of machinery, improves operating results, and provides strong technical support for the realization of refined management of modern agriculture.

[0157] The present invention also provides a storage medium, comprising:

[0158] Memory for storing computer programs;

[0159] A processor is used to implement the above-mentioned method for adjusting the operating parameters of agricultural machinery when executing the computer program.

[0160] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0161] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for adjusting operating parameters of agricultural machinery, characterized in that: The following steps are involved: Step S1: acquiring multispectral image data of farmland through a multispectral camera; performing image segmentation and feature extraction on the multispectral image data, and identifying different soil types, crop growth status, and weed distribution to obtain farmland characteristic map data; using a deep learning algorithm to predict the farming difficulty of different areas on the farmland characteristic map data to obtain farming difficulty prediction data; Step S2: Acquire component image data of agricultural machinery; perform edge detection and morphological analysis on the component image data, and identify the wear state of the component, thereby obtaining wear state data; establish a component wear dynamics model based on a convolutional neural network according to the wear state data, and perform historical wear trend simulation, thereby obtaining historical wear trend data; Step S3: Acquire historical operation log data and current meteorological condition data; perform correlation prediction of operation effect based on historical operation data and current meteorological condition data to obtain operation effect prediction data; perform relationship mapping processing between mechanical operation and component wear on historical operation log data and historical wear trend data to obtain operation-wear mapping data; Step S4: Optimize the operation parameters according to the operation-wear mapping data and the operation effect prediction data, and determine the best parameter combination based on the minimum wear, so as to obtain the optimized operation window data; obtain the operation image stream data in real time through the vehicle-mounted camera system; identify the operation depth and crop coverage rate of the operation image stream data, so as to obtain the real-time operation status data; Step S5: Generate an optimal parameter adjustment strategy based on the farming difficulty prediction data, the real-time operation status data and the optimized operation window data; adjust the agricultural machinery operation parameters in real time according to the optimal parameter adjustment strategy to achieve dynamic adjustment of the operation parameters.

2. The method for adjusting the operating parameters of agricultural machinery according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: deploying a multispectral camera array on the target farmland to obtain multi-band image data of the farmland, wherein the multi-band includes a visible light band, a near infrared band, and a mid-infrared band; Step S12: Preprocessing the multi-band image data based on denoising, contrast enhancement, and illumination effect correction to obtain multispectral image data; Step S13: Calculate the vegetation index and soil index according to the multispectral image data to obtain farmland index map data; Step S14: performing image segmentation and feature extraction on the multispectral image data, and identifying different soil types, crop growth status and weed distribution to obtain farmland characteristic map data; Step S15: using geographic information system technology to perform spatial analysis and interpolation processing on the farmland characteristic map data, thereby obtaining high-precision farmland characteristic map data; Step S16: performing spatial cluster analysis on the high-precision farmland characteristic map data, and dividing farmland areas with similar characteristics, thereby obtaining farmland area division data; Step S17: Input the high-precision farmland feature map data into the pre-trained deep learning model, and predict the farming difficulty of the farmland area division data, so as to obtain the farming difficulty prediction data.

3. The method for adjusting the operating parameters of agricultural machinery according to claim 2, characterized in that: Step S14 includes the following steps: Step S141: performing multi-scale image segmentation on the farmland index map data to obtain preliminary segmentation result data, wherein the multi-scale image segmentation includes small-scale segmentation for identifying small adjustments, medium-scale segmentation for identifying medium-sized features, and large-scale segmentation for identifying large-scale features; Step S142: optimizing the preliminary segmentation result data based on morphological operators to obtain optimized segmentation graph data; Step S143: extracting texture, shape and spectral features based on the optimized segmentation map data, thereby obtaining farmland feature data; Step S144: construct a deep convolutional neural network model based on the pre-acquired labeled sample data, and input the farmland feature data into the deep convolutional neural network model to identify the soil type, crop growth status and weed distribution to obtain farmland feature map data.

4. The method for adjusting the operating parameters of agricultural machinery according to claim 3, characterized in that: Step S2 includes the following steps: Step S21: Acquire component image data of agricultural machinery; Step S22: performing image preprocessing on the component image data to obtain standardized component image data, wherein the image preprocessing includes image denoising, illumination compensation and geometric correction; Step S23: performing multi-level edge detection processing on the standardized component image data and extracting component contour information to obtain component edge feature data; Step S24: extracting component shape features based on morphological analysis of component edge feature data through opening operation, hole filling and edge smoothing, thereby obtaining component morphological feature data; Step S25: comparing the component edge feature data and the component morphology feature data with a pre-established standard component model to identify the wear state of the component and obtain wear state data; Step S26: Establish a component wear dynamics model based on a convolutional neural network according to the wear state data, and perform a historical wear trend simulation to obtain historical wear trend data.

5. The method for adjusting the operating parameters of agricultural machinery according to claim 4, characterized in that: Step S26 includes the following steps: Step S261: construct a convolutional neural network model, the architecture of which includes an input layer, multiple convolutional layers, a pooling layer, a fully connected layer, and an output layer; Step S262: training the convolutional neural network model using the wear state data volume, and optimizing the model parameters by back propagation algorithm to obtain an initial component wear dynamics model; Step S263: using a cross-validation technique to optimize the initial component wear dynamics model by adjusting hyperparameters and increasing or decreasing network layers, thereby obtaining a component wear dynamics model; Step S264: collecting wear status and usage time information under different working conditions, so as to construct a historical wear database; Step S265: Apply the component wear dynamics model to the historical wear database for multiple iterative calculations, simulate the wear progression under different working conditions and time spans, extract wear trend features from the simulation results, and obtain historical wear trend data, where the wear trend features include wear rate, accelerated wear point, and critical wear threshold.

6. The method for adjusting the operating parameters of agricultural machinery according to claim 5, characterized in that: Step S3 includes the following steps: Step S31: Obtain historical operation log data through the operation management system of the agricultural machinery, including operation time, location, operation type, machine parameter settings, operation effect evaluation and meteorological conditions; obtain current meteorological condition data; Step S32: using a machine learning algorithm to perform feature engineering and model training on historical operation data and its corresponding meteorological conditions, thereby obtaining an operation effect prediction model; Step S33: inputting the current meteorological condition data into the trained operation effect prediction model to perform operation effect prediction, thereby obtaining operation effect prediction data; Step S34: performing time alignment and feature matching on the historical operation log data and the historical wear trend data to obtain a time series data set; Step S35: performing relationship mapping processing between mechanical operation and component wear according to the time series data set to obtain operation-wear mapping data.

7. The method for adjusting the operating parameters of agricultural machinery according to claim 6, characterized in that: Step S4 includes the following steps: Step S41: constructing a multi-objective optimization model with minimizing component wear and maximizing operating effect as objective functions according to the operation-wear mapping data and the operation effect prediction data; Step S42: solving the multi-objective optimization model using a genetic algorithm to obtain a series of non-dominated solutions, thereby forming an optimal parameter combination set; Step S43: performing cluster analysis on the optimal parameter combination set to determine the optimal parameter combination interval, thereby generating optimized operation window data; Step S44: collecting image stream data during the operation in real time through the vehicle-mounted camera system; Step S45: using a deep learning image segmentation algorithm to perform semantic segmentation on the image stream data based on the identification of the working area, the non-working area and the crop area, thereby obtaining semantic segmentation result data; Step S46: Calculate the operation depth and crop coverage rate based on the semantic segmentation result data, perform time series analysis on the calculation results, and eliminate instantaneous fluctuations to obtain real-time operation status data, wherein the operation depth and crop coverage rate are calculated by comparing the operation area in the image with the ground reference object, and the crop coverage rate is calculated by the ratio of the crop area to the total area.

8. The method for adjusting the operating parameters of agricultural machinery according to claim 7, characterized in that: Step S5 includes the following steps: Step S51: using the farming difficulty prediction data, the real-time operation status data, and the optimized operation window data as input features to construct a parameter adjustment strategy generation model based on deep reinforcement learning; Step S52: using historical operation data to train the parameter adjustment strategy generation model, thereby obtaining initialization model parameters; Step S53: obtaining current state data; using the initialized model parameters and the parameter adjustment strategy generation model to generate an optimal parameter adjustment strategy according to the current state data; Step S54: adjusting the operating parameters of the agricultural machinery in real time according to the optimal parameter adjustment strategy to achieve dynamic adjustment of the operating parameters.

9. A system for adjusting operating parameters of agricultural machinery, characterized in that: For executing the method for adjusting the operating parameters of agricultural machinery according to claim 1, the system for adjusting the operating parameters of agricultural machinery comprises: The farmland analysis module is used to obtain multispectral image data of farmland through a multispectral camera; perform image segmentation and feature extraction on the multispectral image data, and identify different soil types, crop growth status and weed distribution to obtain farmland characteristic map data; use a deep learning algorithm to predict the difficulty of farming in different areas of the farmland characteristic map data to obtain farming difficulty prediction data; The mechanical diagnosis module is used to obtain the image data of agricultural machinery components; perform edge detection and morphological analysis on the component image data, and identify the wear state of the components to obtain wear state data; establish a component wear dynamics model based on a convolutional neural network according to the wear state data, and perform historical wear trend simulation to obtain historical wear trend data; The operation effect prediction module is used to obtain historical operation log data and current meteorological condition data; perform correlation prediction of operation effects based on historical operation data and current meteorological condition data to obtain operation effect prediction data; perform relationship mapping processing between mechanical operation and component wear on historical operation log data and historical wear trend data to obtain operation-wear mapping data; The parameter optimization module is used to optimize the operation parameters according to the operation-wear mapping data and the operation effect prediction data, and determine the best parameter combination based on the minimum wear, so as to obtain the optimized operation window data; obtain the operation image stream data in real time through the vehicle-mounted camera system; identify the operation depth and crop coverage rate of the operation image stream data, so as to obtain the real-time operation status data; The dynamic adjustment module is used to generate the optimal parameter adjustment strategy based on the farming difficulty prediction data, the real-time operation status data and the optimized operation window data; the agricultural machinery operation parameters are adjusted in real time according to the optimal parameter adjustment strategy to realize the dynamic adjustment of the operation parameters.

10. A storage medium, characterized in that: include: Memory for storing computer programs; A processor, configured to implement the method for adjusting operating parameters of agricultural machinery as described in any one of claims 1 to 8 when executing the computer program.

Images