Satellite remote sensing driven farming operation area intelligent generation management method and system

Abstract

The invention relates to the technical field of satellite remote sensing data processing and geographic space analysis, in particular to a satellite remote sensing driven farm work area intelligent generation management method and system.The method comprises the steps that farm plot boundary vector data is obtained through a geographic information system, and a bounding rectangle is calculated to determine the remote sensing range; downloading satellite data of a sentinel 2; through resampling and cloud removal processing, various growth vigor indexes of crops are calculated, and an in-plot image file is generated according to the growth vigor indexes of the crops; a target index and a threshold value are adaptively selected according to the crop growth stage, a binary image is generated, and opening operation denoising is carried out; and finally, performing connected domain analysis to delete small-area regions, calculating convex hull vertexes of remaining regions, converting the convex hull vertexes into longitudes and latitudes, and obtaining intersections with plot boundaries to generate operation region boundary data. After correction, the information is issued to the intelligent agricultural machine terminal, closed-loop management is completed, precise generation and efficient management of the operation area are achieved, and the requirement for precise operation of modern agriculture is met.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of satellite remote sensing data processing and geographic spatial analysis, in particular to a satellite remote sensing driven intelligent generation management method and system for agricultural operation area. BACKGROUND

[0002] Satellite remote sensing data processing and geographic spatial analysis technology is an important technology, which is specifically applied to the precise planning and closed-loop management link of agricultural operation area of large-scale farms, and the core is to realize crop growth monitoring, intelligent generation of operation area and operation of agricultural machinery through the fusion of satellite remote sensing technology and spatial algorithm, so as to adapt to the demand of modern agriculture for precision, high efficiency and traceable agricultural management; The current agricultural operation area planning related technology has significant limitations, which is difficult to meet the precise management needs of large-scale farms. The existing technology does not establish a precise adaptation mechanism between farm plot boundaries and satellite remote sensing data. Directly downloading fixed range remote sensing data is easy to cause data redundancy or incomplete plot coverage, and lacks targeted band fusion and cloud removal processing, which leads to interference in crop growth index calculation, insufficient data reliability, and dynamic selection of adaptive growth evaluation index according to different growth stages of crops. The threshold value is mostly a fixed value, which cannot adapt to the different growth characteristics of the seedling stage, growth period and mature period, resulting in large deviation of operation area identification. In the image processing process, the burr and noise are not effectively eliminated, and the small area invalid region is not filtered, which may cause misjudgment in connected domain analysis, affecting the integrity and accuracy of the operation area boundary. There is no precise conversion of pixel coordinates to geographic latitude and longitude and spatial overlay analysis link, and the generated operation area may exceed the farm plot range, which has poor practicability. After the generation of the operation area, it is difficult to correct according to the actual scene, and it is also difficult to efficiently synchronize to the intelligent agricultural machinery terminal, resulting in low management efficiency and serious resource waste. It cannot meet the needs of modern agricultural precision and efficient agricultural operation management. In order to solve this technical problem, we provide a satellite remote sensing driven intelligent generation management method and system for agricultural operation area. SUMMARY

[0003] The present application aims to provide a satellite remote sensing driven intelligent generation management method and system for agricultural operation area to solve the problems raised in the background art.

[0004] To achieve the above-mentioned purpose, one of the purposes of the present application is to provide a satellite remote sensing driven intelligent generation management method for agricultural operation area, comprising the following steps: S1, obtaining farm plot boundary vector data, calculating the circumscribed rectangle of all plot boundary sets to determine the geographic range of satellite remote sensing data; S2, download Sentinel 2 satellite remote sensing data, perform band fusion and cloud removal processing to calculate crop vigor index, generate an image file according to the crop vigor index, and then use the farm plot boundary vector data to crop the image file to generate a cropped image file; S3, select a target index based on the current growth stage of the crop and set a threshold, read the cropped image file, extract a pixel matrix, and combine the pixel matrix and the set threshold to generate a binary image, wherein the pixel value less than the threshold is set to 1 to represent the required operation area, otherwise set to 0, and perform an open operation on the binary image to eliminate burrs and noise; S4, traverse the binary image to perform connected component analysis, delete connected components with an area less than a preset area, calculate the convex hull vertex pixel coordinates of the remaining connected components, and convert the convex hull vertex pixel coordinates to latitude and longitude coordinates, using the metadata in the image file, including the top-left latitude and longitude coordinates and the resolution parameter, combine the convex hull vertex latitude and longitude coordinates into a MultiPolygon object, and combine all the plot boundaries of the farm into a MultiPolygon object, calculate the intersection of the two MultiPolygon objects, generate operation area boundary data and store it in a spatial database; S5, finally, read the operation area boundary data from the spatial database, and allow correction according to actual conditions, after correction, the final operation range is issued to the intelligent agricultural terminal to perform agricultural operation, completing the closed-loop management process.

[0005] The second object of the application is to provide a system for implementing a satellite remote sensing driven agricultural operation area intelligent generation management method comprising any one of the above, comprising: The data preprocessing unit is used to obtain farm plot boundary vector data and calculate the circumscribed rectangle of the union of all plot boundaries to determine the geographic range of satellite remote sensing data, download Sentinel 2 satellite remote sensing data and perform resampling and cloud removal processing, calculate the crop vigor index and generate an image file according to the index, and then use the farm plot boundary vector data to crop the image file to generate a cropped image file; The intelligent analysis unit is used to adaptively select a target index based on the current growth stage of the crop and set a threshold, read the cropped image file to extract a pixel matrix, combine the threshold to generate a binary image and perform an open operation on the binary image to eliminate burrs and noise, traverse the binary image to perform connected component analysis and delete connected components with an area less than a preset area, and calculate the convex hull vertex pixel coordinates of the remaining connected components; The spatial optimization unit is used to convert the convex hull vertex pixel coordinates to latitude and longitude coordinates and combine them into a MultiPolygon object, combine all the plot boundaries of the farm into a MultiPolygon object, calculate the intersection of the two MultiPolygon objects to generate operation area boundary data and store it in a spatial database; The closed loop execution unit is used to read the job area boundary data from the spatial database for the manager to correct, and after correction, the final job range is issued to the intelligent agricultural machine terminal through the wireless communication protocol, the agricultural work is executed, and the closed loop management process is completed.

[0006] Compared with the prior art, the present application has the following advantages: The present application calculates the circumscribed rectangle of the farm plot through the data preprocessing unit, accurately locks the satellite remote sensing range, avoids data redundancy or incomplete coverage, and improves the data quality through band fusion and cloud removal processing, ensures the reliability of the crop growth index calculation, solves the problems of poor data adaptation and interference in the traditional data, the intelligent analysis unit dynamically selects the adaptive index according to the crop growth stage, sets the threshold value based on the historical data, combines the opening operation to remove noise and filter small areas in the connected domain, greatly improves the identification accuracy of the work area, avoids the deviation caused by the fixed index threshold, the spatial optimization unit accurately converts the pixel and latitude and longitude coordinates through coordinate affine transformation, superimposes and analyzes to obtain the intersection of MultiPolygon objects, ensures that the work area is within the farm plot, solves the problem of inaccurate boundary, the closed loop execution unit supports the manager to correct the data and issue it to the intelligent agricultural machine terminal, forms a closed loop, improves the efficiency of agricultural management, and fully meets the needs of modern agricultural precision and efficient operation management. BRIEF DESCRIPTION OF DRAWINGS

[0007] Figure 1 is the overall work flowchart of the present application; Figure 2 is the overall structure schematic diagram of the present application; The meanings of various reference numerals in the drawings are as follows: 1, data preprocessing unit; 2, intelligent analysis unit; 3, spatial optimization unit; 4, closed loop execution unit. DETAILED DESCRIPTION

[0008] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the present application.

[0009] Please refer to Figure 1 The present embodiment aims to provide a satellite remote sensing driven agricultural work area intelligent generation management method, which comprises the following steps: S1, obtain the farm plot boundary vector data, calculate the circumscribed rectangle of the union of all plot boundaries to determine the geographical range of satellite remote sensing data; S2, download Sentinel-2 satellite remote sensing data, perform band fusion and cloud removal processing to calculate multiple crop vigor indexes, generate image files according to the crop vigor indexes, and use the farm plot boundary vector data to crop the image files to generate cropped image files; S3, select a target index based on the current growth stage of the crop and set a threshold, read the cropped image files, extract the pixel matrix, and combine the pixel matrix and the set threshold to generate a binary image, where the pixel value less than the threshold is set to 1 to represent the area that needs to be operated, otherwise set to 0, and perform an open operation on the binary image to eliminate burrs and noise; S4, traverse the binary image to perform connected component analysis, delete connected components with an area less than a predetermined area, calculate the convex hull vertex pixel coordinates of the remaining connected components, and convert the convex hull vertex pixel coordinates to latitude and longitude coordinates, using the metadata in the image file, including the top-left latitude and longitude coordinates and the resolution parameters, combine the convex hull vertex pixel coordinates into a MultiPolygon object, and combine all the farm plot boundaries into a MultiPolygon object, calculate the intersection of the two MultiPolygon objects, generate the operation area boundary data and store it in the spatial database; S5, finally, read the operation area boundary data from the spatial database and allow it to be modified according to actual conditions, after modification, the final operation range is issued to the intelligent agricultural terminal to perform agricultural operations, completing the closed-loop management process.

[0010] The farm plot boundary vector data is obtained by importing the pre-stored farm vector data through a geographic information system, which contains a single plot boundary coordinate sequence, and calculates the circumscribed rectangle of the plot boundary set.

[0011] The downloaded Sentinel-2 satellite remote sensing data is based on a public data interface to automatically obtain multi-spectral band data, resample different resolution band data to the same resolution, use CLDPRB band data and a pre-set threshold to generate a cloud coverage mask, and set the values of other bands covered by clouds to zero.

[0012] When calculating the crop vigor index, generate multiple crop vigor indexes according to the pre-set index calculation formula and directly output as an image file, then use the farm plot boundary vector data to crop the image file, and only keep the pixel area within the plot by vector and raster overlay analysis to generate the cropped image file.

[0013] Selecting a target index based on the current growth stage of the crop and setting a threshold, specifically includes: The growth stage is divided into the emergence period, growth period and mature period by dynamic matching of the growth model. The index type sensitive to seedling density is selected in the emergence period, the index type sensitive to canopy coverage is selected in the growth period, and the index type sensitive to nutrient status is selected in the mature period. The threshold value is adaptively generated according to the classification model trained according to historical data. After reading the cut image file, the index value matrix is extracted.

[0014] When the pixel matrix and the set threshold value are combined to generate a binary image, the index value and the threshold value are compared pixel by pixel. When the index value is less than the threshold value, the area needing operation is marked and assigned a value of 1, otherwise the non-operation area is marked and assigned a value of 0. Morphological filtering is used for open operation of the binary image. First, erosion is performed, and then inflation is performed to eliminate burrs and noise.

[0015] When the binary image is traversed for connected domain analysis, the region growing algorithm is used to identify adjacent pixel regions, and the actual area of each connected domain is calculated. The connected domains with an area less than the preset area are deleted by area threshold filtering. The preset area is dynamically adjusted according to the farm scale. The convex hull vertex pixel coordinates of the remaining connected domains are calculated one by one using the Sklansky algorithm. The convex hull vertex is solved using the connected domain boundary point pixel coordinates as input.

[0016] When the convex hull vertex pixel coordinates are converted into latitude and longitude coordinates, the metadata in the image file including the top-left latitude and longitude coordinates and the resolution parameters are directly read. The conversion process is realized by coordinate affine transformation to map the pixel coordinates to the geographic coordinates.

[0017] When the convex hull vertex latitude and longitude coordinates are combined into a MultiPolygon object, the latitude and longitude coordinate sequence converted from a single convex hull is first packaged into a single Polygon object in counterclockwise order, and then multiple Polygon objects are combined into a MultiPolygon object. When the farm plot boundaries are combined into a MultiPolygon object, a single Polygon object is first packaged from a single plot boundary, and then the Polygon objects of all plots in the farm are combined into a MutiPolygon object.

[0018] When the intersection of two MultiPolygon objects is calculated, the spatial overlay analysis algorithm is used to calculate the intersection area of the convex hull MultiPolygon object and the plot boundary MultiPolygon object, generate the operation area boundary data, and store the data in the spatial database using the geometry type field.

[0019] It needs to be further explained that as the basic link of satellite remote sensing driving agricultural operation area generation, accurate acquisition of farmland plot boundary data and determination of remote sensing data geographic range are the premise of subsequent data download and processing. Only by clearly defining the spatial distribution and bounding range of the plot, can effective remote sensing data be targetedly acquired, and data redundancy or incomplete coverage be avoided. The specific implementation is as follows: The farmland plot boundary vector data refers to the geographic spatial data characterized by the coordinate sequence form of the plot outline. Its core is a closed polygon formed by a series of latitude and longitude coordinates, which can accurately reflect the actual boundary of a single plot. The data is obtained by geographic information system (GIS). The geographic information system (GIS) is a professional software system with the functions of geographic data import, management and analysis. Here, the built-in data import module of the system is used to read the farmland vector data pre-stored in the system database. These pre-stored data are input by the farm managers in advance or obtained through field surveying and mapping. Each piece of data corresponds to a single plot and contains a complete plot unique identifier and boundary coordinate sequence. The boundary coordinate sequence is a set of latitude and longitude coordinates arranged in clockwise or counterclockwise order, which ensures that the plot outline can be completely outlined.

[0020] During the import process, the GIS system automatically checks the integrity and closure of the coordinate sequence. If the coordinate sequence of a plot does not form a closed polygon, a prompt will be displayed and the administrator will be allowed to manually correct it until all plot boundary data meet the spatial analysis requirements. After the acquisition of plot boundary data is completed, the next step is to calculate the bounding rectangle of the collection of all plot boundaries to determine the geographic range of satellite remote sensing data. The bounding rectangle refers to the smallest rectangle that can completely contain all plot boundaries. Its four sides are parallel to the meridians and latitudes. Through the bounding rectangle, the latitude and longitude intervals that the remote sensing data needs to cover can be determined, avoiding the download of redundant data. When calculating, the boundary coordinate sequence of all plots is taken as the unified input, that is, all plot coordinate points are integrated into a global point set. The four key points corresponding to the maximum x-axis (longitude) value, the minimum x-axis value, the maximum y-axis (latitude) value and the minimum y-axis value are selected, and the four points are taken as the vertices of the bounding rectangle.

[0021] After the geographic range of remote sensing data is determined, the acquisition and preprocessing of Sentinel-2 satellite remote sensing data can be started. The core of this process is to obtain high-quality, interference-free multispectral data to provide reliable data sources for subsequent crop vigor index calculation. The specific implementation is as follows: The Sentinel 2 satellite remote sensing data is multispectral remote sensing data obtained by the Sentinel 2 satellite launched by the European Space Agency (ESA), which contains multiple bands such as visible light and near-infrared, and can effectively reflect the crop growth state. Data download is based on the public data interface of the Sentinel satellite. The system calls the interface through programming, inputs the latitude and longitude range of the calculated bounding rectangle, the data acquisition time (usually select a period of time within 7 days without extreme weather) and the required band type, the interface will automatically match and return the remote sensing data file that meets the conditions. During the download process, the system will monitor the data integrity in real time. If the network is interrupted and the download fails, it will automatically re-initiate the request to ensure complete data acquisition. After the download is completed, the data of different resolution bands are unified to 10m resolution through resampling. After resampling, the cloud coverage mask is generated using the CLDPRB band data and the preset threshold, and the values of the parts covered by the cloud in other bands are set to zero.

[0022] The remote sensing data after resampling and cloud removal has the basic conditions for calculating the crop vigor index. The crop vigor index can quantitatively reflect the growth status of the crop, and the targeted cutting of the image can focus on the farm plot range and eliminate irrelevant area interference. The specific implementation is as follows: The crop growth index is a quantitative index calculated based on the reflectivity of the remote sensing data band, which can indirectly reflect the growth information of the crop biomass, canopy coverage, etc. The core calculation can reflect the index of crop growth, such as NDVI, SAVI, RVI, EVI, etc. Taking the NDVI index as an example, the normalized difference vegetation index (NDVI) is the most widely used crop growth evaluation index. The core logic is to use the difference between the high reflectivity of the near-infrared band and the low reflectivity of the red light band of the crop, and convert it into a quantitative value through a preset formula. During calculation, the system calls the preprocessed remote sensing image data, extracts the near-infrared band and red light band reflectivity value of each pixel, and substitutes it into the preset index calculation formula to generate the equivalent characteristic value of the normalized vegetation index corresponding to each pixel. The larger the value, the better the crop growth, otherwise the growth is weak, and agricultural operations such as fertilization and irrigation may be needed. After generating the growth index values of all pixels, the image file is directly output. The image file refers to a raster-formatted remote sensing image file, each pixel corresponds to a unique growth index value, and the original remote sensing data metadata is retained for subsequent spatial coordinate conversion. Then, using the obtained farm plot boundary vector data, the image file is cut. The core of the cutting process is to retain only the pixel area within the plot boundary through vector and raster overlay analysis, and exclude irrelevant areas outside the plot. Vector and raster overlay analysis is a core technology of geographic spatial analysis. In specific implementation, first, align the farm plot boundary vector data and the growth index image file in space to ensure that their projection coordinate systems and pixel resolutions are completely consistent. Then, traverse each pixel of the image file to determine whether the center coordinates of the pixel fall within any plot boundary polygon. If the pixel center coordinates are within the plot boundary, retain the growth index value of the pixel. If it is outside the boundary, set the pixel value to an invalid value. After traversal, the cut image file is generated. This file only contains the growth index information within the farm plot, which not only reduces the data volume for subsequent analysis, but also avoids the interference of irrelevant areas on the operation area identification, laying the foundation for the next step of binary image generation.

[0023] After completing the cutting of the growth index image file, the target index for the current growth stage of the crop needs to be selected accurately and the threshold value needs to be set. The growth needs and sensitive features of crops at different growth stages are significantly different. If a single index and a fixed threshold are used uniformly, it will cause obvious deviation in the identification of the operation area. Therefore, the growth stage needs to be dynamically matched through the growth model, and the adaptive threshold value needs to be generated relying on the classification model trained based on historical data to ensure that the index and the threshold value can accurately match the actual evaluation needs of the current growth stage. The specific implementation is as follows: Firstly, the current growth stage of the crop is dynamically matched through a growth model, which is a quantitative simulation model based on the physiological rules of crop growth and environmental factors. Here, an accumulated effective temperature (the sum of daily average temperatures above the biological zero required for crop growth and development) driven growth model is used. The core divides the crop growth cycle into three core stages: the emergence period, the growth period, and the mature period. The emergence period is from sowing to the emergence of seedlings and the unfolding of the first true leaf, usually corresponding to an accumulated effective temperature of 50-80 °C·d. The growth period is the period of vegetative growth after emergence to the onset of reproductive growth, corresponding to an accumulated effective temperature of 80-300 °C·d. During this period, the crop canopy expands rapidly and biomass increases significantly. The mature period is from flowering and fruiting to grain maturity, corresponding to an accumulated effective temperature of more than 300 °C·d. During this period, crop growth slows down, and the focus is on nutrient accumulation and water supply in the later stages. The growth model reads the daily average temperature data from the local weather station in real time, calculates the accumulated effective temperature from the sowing day, and then determines the current growth stage through dynamic matching. If there is a short-term missing of accumulated effective temperature data, it is interpolated by the average accumulated effective temperature of the previous 5 days to ensure the continuity of the growth stage determination. Based on the determined growth stage, the corresponding target index is selected, and the core logic is the matching of stage requirements and index sensitivity. In the emergence period, the crop focuses on seedling emergence and density establishment, and a density-sensitive index type needs to be selected. Here, the ratio vegetation index (RVI) is preferred. This index calculates the ratio of near-infrared band reflectance to red band reflectance, and is more sensitive to low-coverage seedling areas, allowing accurate differentiation between seedling sparse areas requiring reseeding and areas with adequate density. In the growth period, the crop canopy coverage increases rapidly, and a canopy coverage-sensitive index type needs to be selected. At this time, the enhanced vegetation index (EVI) is selected. This index introduces a soil adjustment factor and an atmospheric correction term, effectively reducing the interference of soil background and atmospheric scattering, and can more accurately reflect the dynamic changes of canopy coverage. In the mature period, nutrient reserves and water status directly affect yield, and a nutrient status-sensitive index type needs to be selected. Here, the soil-adjusted vegetation index (SAVI) is selected. This index adds a soil brightness correction coefficient to reduce the influence of soil background on the index value, and can better associate with the nutrient accumulation of crops. The threshold setting is generated adaptively by a classification model trained on historical data. The classification model trained on historical data refers to the growth data of the same crop and the same plot in the past 3 to 5 years as samples. A prediction model is constructed through machine learning algorithms, and the sample data includes growth stage, target index value, actual operation area label, and yield data. For example, in the emergence period of wheat, the sample records the RVI value, whether reseeding operation is performed (actual operation area label), and the final yield per mu of the area. The model learns these data to establish the mapping relationship between growth stage, environmental factors, target index value, and whether operation is needed.

[0024] In the current threshold generation, the model inputs the current growth stage, real-time environmental data, and the index value statistical characteristics of the cropped image, and adaptively outputs the corresponding threshold. For example, in the seedling stage with high soil moisture, the model may generate an RVI threshold of 0.75 (as the seedlings grow slightly faster in high humidity, the threshold can be appropriately lowered), while in the seedling stage with dry soil, the threshold may be raised to 0.85 (dryness causes slow growth of seedlings, and a more stringent determination of sparse areas is needed), avoiding the problem of fixed threshold unable to adapt to environmental differences.

[0025] After completing the target index selection and threshold setting, the cropped image file is read and the pixel matrix is extracted. This process is achieved through a raster analysis tool, which is a professional tool with raster data reading and analysis functions. Its core function is to convert TIFF format index image files into two-dimensional pixel matrices that can be used for numerical operations. In specific implementation, the raster analysis tool first reads the metadata of the image file to determine the dimensions of the matrix, and then reads the index values pixel by pixel from left to right and from top to bottom. The target index value of each pixel is assigned to the corresponding element of the matrix. During the reading process, the tool automatically skips invalid pixels and marks them with a specific symbol in the matrix, finally generating a complete pixel matrix, providing structured numerical input for subsequent binary image generation.

[0026] After obtaining the structured pixel matrix and adaptive threshold, the index information needs to be converted into intuitive work area markers (binary images) through numerical comparison. However, the directly generated binary image is easily disturbed by image noise and burrs, leading to misjudgment in subsequent connected component analysis. Therefore, it is necessary to first generate a binary image through comparison operation, and then rely on morphological filtering open operation to optimize image quality. The specific implementation is as follows: First, the pixel matrix is compared with the set threshold value, and a binary image is generated. The binary image refers to an image containing only two pixel values (usually 0 and 1), where 1 represents the region to be worked on, and 0 represents the non-working region. The core is to convert the abstract long potential index into a concrete region label through the quantitative comparison of the index value and the threshold value. In specific implementation, the system will traverse each valid element in the pixel matrix, and perform a comparison operation on each pixel. The target index value of the current pixel is compared with the adaptively generated threshold value. If the index value is less than the threshold value, it means that the crop growth potential of the pixel corresponding to the land region does not meet the expected value, and agricultural work is needed. Therefore, the pixel value is assigned as 1. If the index value is greater than or equal to the threshold value, it means that the crop growth potential of the region meets the standard, and no work is needed. The pixel value is assigned as 0. For example, in the wheat growing period, the EVI threshold value is set to 1.2, and the EVI value of a certain pixel is 1.1 (less than the threshold value), which is assigned as 1 (fertilization is needed). The EVI value of another pixel is 1.3 (greater than the threshold value), which is assigned as 0 (no fertilization is needed). After traversal, the pixel matrix originally containing continuous index values is converted into a binary image containing only 0 and 1. At this time, the region to be worked on is presented in the form of a 1-pixel cluster, and the non-working region is filled with 0, directly reflecting the preliminary distribution of the working region. However, the initial binary image generated often contains burrs and noise. Burr refers to the fine isolated pixel chain on the edge of the working region, and noise refers to a single or small number of discrete 1-pixels. These disturbances may cause the subsequent connected component analysis to mistakenly identify them as independent working regions, increasing the invalid path of agricultural machinery work. Therefore, it is necessary to eliminate them through the opening operation of morphological filtering. Morphological filtering is an algorithm based on image morphological features. The opening operation is one of its core operations, which involves a two-step processing flow of erosion followed by dilation. Through this process, the main working region can be preserved while removing fine burrs and isolated noise.

[0027] The first step performs the erosion operation. Erosion refers to the operation of reducing the working region (1-pixel region) by a predetermined structure element traversing the binary image according to the rules. A 3x3 square structure element is selected here, which is a 3-pixel x 3-pixel matrix with the current pixel at the center and 8 neighboring pixels around it. When the structure element traverses the image, the center pixel is only retained as 1 if all 9 pixels covered by the structure element are 1. If any of the pixels is 0, the center pixel is assigned as 0. For example, for a single isolated 1-pixel (noise), its surrounding pixels are all 0. When the structure element traverses, the center pixel will be assigned as 0, achieving noise elimination. For a 1-pixel wide burr region, the structure element will cover the pixels of the burr region one by one, and due to the existence of 0 pixels in the neighborhood, the pixels of the burr region will be assigned as 0 one by one, achieving burr removal. After the erosion operation, the main working region will be slightly reduced, but the core outline will be preserved.

[0028] The second step performs an expansion operation, which is the inverse of the erosion operation. The image is traversed using a 3x3 square structuring element, and the rule is: if any of the 9 pixels covered by the structuring element is 1, then the center pixel is assigned a value of 1. The purpose of this operation is to restore the area of the main work area that was reduced during the erosion process. For example, a large area of continuous 1-pixel clusters will have their edges reduced after erosion. During expansion, the 1 values of the neighboring pixels will reassign the edge pixels to 1, restoring them to near their original size. Spikes and noise that have been eliminated by erosion cannot be restored after expansion because there are no continuous 1-pixels around them, thus achieving the effect of removing noise and spikes while preserving the effective area.

[0029] Through the opening operation of erosion followed by expansion, an optimized binary image is finally generated. In this image, only the larger and complete outline work areas (1-pixel clusters) are retained, and the fine spikes and isolated noise are completely eliminated, laying a clear image foundation for subsequent image traversal and accurate extraction of work area boundaries.

[0030] After completing the opening operation optimization of the binary image, although the spikes and isolated noise have been eliminated, there may still be small work areas. If these areas are included in the work range, it will increase the number of turns and invalid paths of agricultural machinery, reducing the work efficiency. Therefore, it is necessary to identify the effective area through connected component analysis and filter small areas, and then optimize the region boundary through the convex hull algorithm. The specific implementation is as follows: First, traverse the optimized binary image and perform connected component analysis using the region growing algorithm. A connected component is a set of adjacent pixels in a binary image that have the same value (1 in this case, representing the work area). The region growing algorithm starts with a single seed pixel and gradually absorbs adjacent pixels with the same value to form a connected component. In implementation, first initialize a label matrix with the same size as the binary image, with all initial values being 0, which is used to label the connected component number of each pixel. Then traverse the binary image from left to right and top to bottom. When a pixel with a value of 1 is encountered and the corresponding position in the label matrix is 0, set it as a seed pixel and assign it a unique connected component number. Then use 8-neighborhood search, which searches for 8 adjacent pixels above, below, left, right, and diagonally from the seed pixel. This can more completely capture pixels connected diagonally than 4-neighborhood search, avoiding missing connected regions. All adjacent pixels with a value of 1 that have not been labeled are included in the current connected component, and the label matrix number is updated. Repeat the neighborhood expansion step until no new same-value pixels can be included. At this time, the current connected component is complete. Continue to traverse the image to find the next unlabeled seed pixel and repeat the above process until all pixels with a value of 1 are labeled as different connected components.

[0031] After the connected domain recognition is completed, the actual area of each connected domain is calculated, and the actual area refers to the actual area of the ground corresponding to the connected domain, rather than the number of pixels, and the calculation needs to be combined with the resolution parameter of the remote sensing image. The specific steps are as follows: first, the total number of pixels contained in each connected domain is counted (by traversing the marking matrix, the number of pixels with the same number is counted), and then the actual area of the connected domain is obtained by multiplying the total number of pixels by the ground area of a single pixel; then, through area threshold filtering, the connected domains with an area less than the preset area are deleted, and the preset area is a critical value dynamically adjusted according to the operation requirements, and the core is to avoid the interference of small areas to the operation planning. When filtering, the actual area of each connected domain is compared with the preset area of the corresponding gear, and if the actual area is less than the preset area, the pixel value of the connected domain in the marking matrix is reset to 0, which is regarded as a non-operation area, and only the areas that meet the standard are retained.

[0032] For the remaining connected domains that meet the area standard, the convex hull vertex pixel coordinates are calculated, and this process is realized by Andrew algorithm. Andrew algorithm is an algorithm for finding the smallest convex polygon (i.e. convex hull) that can completely contain a set of points, and the convex hull vertex is the vertex of the smallest convex polygon. By convex hull, the irregular connected domain boundary can be optimized to a simple polygon contour, which is convenient for subsequent coordinate conversion and operation area boundary determination. In specific implementation, first, the boundary point set of each connected domain that meets the standard is extracted, and the pixels of the connected domain are traversed to select the pixels located at the edge of the region, i.e. the pixels with at least one adjacent pixel value of 0. The coordinates (row number, column number) of these pixels constitute the boundary point set, and then the boundary point set is input into Andrew algorithm, and the algorithm steps are as follows: Find the point with the smallest x-coordinate (the leftmost point) and the point with the largest x-coordinate (the rightmost point) in the boundary point set, and these two points are the selected vertices of the convex hull. At the same time, the point set is divided into upper and lower parts. First, build the upper convex shell: traverse the points in the upper part from the leftmost point to the rightmost point, and retain the points that can make the three-point line in counterclockwise direction, and remove the points in the concave part to form the upper convex shell vertex sequence. Then build the lower convex shell: traverse the points in the lower part from the rightmost point to the leftmost point, and also retain the points in counterclockwise direction to form the lower convex shell vertex sequence. Finally, combine the vertex sequences of the upper and lower convex shells, remove the duplicate vertices, and get the convex hull vertex pixel coordinates of the connected domain, which are arranged in clockwise or counterclockwise order.

[0033] After obtaining the convex hull vertex pixel coordinates of the remaining connected domains, these coordinates only represent the pixel positions in the image and cannot be directly used for spatial positioning of agricultural machinery operation. Agricultural machinery operation needs to rely on geographic coordinates (latitude and longitude) to determine the actual plot position, so it is necessary to convert the pixel coordinates into latitude and longitude coordinates, which needs to be realized by means of the metadata of the image file and coordinate affine transformation. The specific implementation is as follows: Firstly, the metadata in the image file is read, which refers to auxiliary data describing the spatial information of remote sensing images, shooting parameters, etc. Here, two types of key information are extracted. One is the latitude and longitude coordinates of the upper left corner, which is the ground geographical coordinates corresponding to the first pixel in the upper left corner of the image, usually using the WGS84 coordinate system. The second is the resolution parameter, which is the actual distance on the ground corresponding to a single pixel, with a unit of meters. The native resolution of the multispectral band of Sentinel 2 satellite is 10 meters, and some bands are 20 meters. Here, it is converted to 10-meter resolution to ensure accuracy, which means that the distance between two adjacent pixels on the ground is 10 meters, and the east-west direction and the north-south direction are both 10 meters. Metadata is usually stored in the header information of the image file and can be directly read through geographic data processing libraries such as GDAL. After reading, the correctness of the latitude and longitude format needs to be verified. If there is a format exception, the coordinate calibration tool needs to be called for correction. The conversion process is achieved through coordinate affine transformation, which is a mathematical method of mapping pixel coordinates (image coordinate system) to geographical coordinates (latitude and longitude coordinate system) through linear transformation (translation, scaling). The core is to establish the corresponding relationship between pixel row and column numbers and latitude and longitude. First, define the two coordinate systems: In the image coordinate system, the pixel coordinates take the upper left corner of the image as the origin (0, 0), the x-axis along the horizontal direction to the right (corresponding to the column number of the image, denoted as col), and the y-axis along the vertical direction downward (corresponding to the row number of the image, denoted as row). In the geographical coordinate system, the longitude increases along the east-west direction (eastward is positive), and the latitude increases along the north-south direction (northward is positive). The specific transformation calculation logic is as follows: For the pixel coordinates (row, col) of a convex hull vertex, the corresponding longitude is calculated as the longitude of the upper left corner + col x longitude increment, where the longitude increment is the longitude change value corresponding to a single pixel in the east-west direction. Because the earth is a sphere, the ground distance corresponding to 1 degree of longitude is different at different latitudes. Here, an approximate calculation is used, taking 1 degree of longitude at the equator to be approximately equal to 111,000 meters, so the longitude increment = resolution (meters) / 111,000 (meters / degree). For example, at a resolution of 10 meters, the longitude increment = 10 / 111,000 ≈ 0.00009 degrees / pixel. The latitude is calculated as the latitude of the upper left corner - row x latitude increment, where the latitude increment is the latitude change value corresponding to a single pixel in the north-south direction, and 1 degree of latitude corresponds to a ground distance of approximately 111,000 meters, which is approximately constant globally. Thus, the latitude increment = resolution / 111,000. Again, using a resolution of 10 meters as an example, the latitude increment is approximately equal to 0.00009 degrees / pixel. The reason for using subtraction is that the y-axis in the image coordinate system is downward (corresponding to the south direction on the ground), and the latitude decreases in the south direction. Therefore, the larger the row (the closer to the bottom of the image), the smaller the latitude value, and the latitude value needs to be subtracted from the latitude of the upper left corner by the product of row and the latitude increment. For example, for a convex hull vertex with pixel coordinates (row = 20, col = 30) and an image upper left corner longitude and latitude of (116.325°E, 39.938°N), at a resolution of 10 meters: longitude = 116.325 + 30 x 0.00009 ≈ 116.3277°E; latitude = 39.938 - 20 x 0.00009 ≈ 39.9362°N. Thus, the longitude and latitude coordinates of the vertex are obtained, and all convex hull vertices are calculated in this way to generate a complete sequence of convex hull vertex longitude and latitude coordinates, providing geographic spatial data for subsequent combination of MultiPolygon objects.

[0034] After completing the conversion of convex hull vertex pixel coordinates to longitude and latitude coordinates, these discrete coordinate sequences need to be packaged into structured MultiPolygon objects. MultiPolygon objects are the basic data format for geographic spatial analysis, and can accurately represent the boundary range of the work area and the farm plot. By merging the plot boundaries to generate a unified MultiPolygon object, the foundation is laid for subsequent calculations of the intersection and determination of the final work area. The specific implementation is as follows: First, the longitude and latitude coordinate sequence of a single convex hull vertex is encapsulated into a Polygon object, which refers to a geographic geometric entity composed of a closed longitude and latitude coordinate sequence. It needs to meet the requirements of the coordinate sequence arranged in a specific direction and the first and last coordinates consistent. Here, the coordinates are arranged in a counterclockwise order, which is the general standard of geographic spatial data. Its core role is to clearly define the outer boundary of the polygon. The area enclosed in the counterclockwise direction is the interior of the polygon, and the clockwise direction is the exterior. This avoids confusion between the inner and outer boundaries during subsequent spatial overlay analysis. In specific implementation, first, arrange the longitude and latitude coordinate sequence of the convex hull vertex in a counterclockwise order: calculate the average longitude and latitude of all vertices (center point coordinates), then calculate the azimuth angle of each vertex and the center point, rotate the angle from the north direction of the center point clockwise to the angle of the connecting line, rearrange the coordinate sequence in the order of azimuth angle from small to large (corresponding to counterclockwise direction); after sorting, check whether the first and last coordinates of the sequence are consistent. If not, add the first coordinate as the last coordinate to the end of the sequence to ensure the closure of the polygon; finally, through the Polygon class of the geographic data processing library, encapsulate the sorted closed coordinate sequence into a Polygon object. This object contains the area, boundary length, and other attributes of the polygon, and can be directly used for subsequent spatial operations. Then, all Polygon objects of the convex hull are combined into a MultiPolygon object.

[0035] When combining the boundaries of the farm plots into a MultiPolygon object, first encapsulate the boundary of a single plot into a single Polygon object, and then combine the Polygon objects of all plots in the farm into a MultiPolygon object. This process is achieved by merging plot vector data, which refers to the single plot boundary data pre-stored in the geographic information system, and each piece of data is a vector record containing plot ID and closed longitude and latitude coordinate sequence. When merging, the problem of adjacent plot common edges needs to be handled. If two plots are adjacent, there will be repeated coordinate sequences (common edges) in their boundaries. Direct merging will cause gaps or overlaps in the MultiPolygon. Therefore, the coordinate sequences of all plots need to be de-duplicated and spliced first. Traverse the coordinate sequences of all plots, extract the boundary point set of each plot, identify the common edges of adjacent plots through spatial topology analysis, that is, the continuous coordinate fragments that are exactly the same in two plots, delete the common edge coordinates of one of the plots, and only keep one common edge, then merge the two plots into a polygon; for non-adjacent plots, keep their complete coordinate sequences. For example, a farm contains two adjacent plots A and B, the boundary sequence of plot A contains common edge coordinates (P1-P2-P3), and the sequence of plot B also contains this common edge. When merging, delete (P1-P2-P3) of plot B, and only keep the fragment of plot A. Then, splice the remaining sequence of plot A with the remaining sequence of plot B in counterclockwise order to form a unified coordinate sequence covering plots A and B. If the farm contains more plots, repeat the above steps of processing common edges and splicing sequences, and finally generate a closed coordinate sequence covering all plots. Similarly, encapsulate the de-duplicated and spliced boundary of a single plot into a Polygon object through the Polygon class. Then, combine the Polygon objects of all plots in the farm into a MultiPolygon object.

[0036] Through the above process, two core MultiPolygon objects are finally obtained, one is the candidate MultiPolygon of the area to be operated composed of the convex hull vertices, and the other is the total MultiPolygon of the farm plots composed of the merged boundaries of all plots. Both of them are standard geographic geometric objects and use the same coordinate system, which provides a unified data basis for subsequent spatial overlay analysis to calculate the intersection and accurately extract the area to be operated within the farm plots, ensuring the spatial accuracy of the subsequent operation area boundary data.

[0037] After successfully generating the convex hull MultiPolygon object (the candidate range of the work area) and the total MultiPolygon object of the farm plot (the actual boundary of the farm plot), the effective work area needs to be accurately locked through intersection operation. The convex hull MultiPolygon can optimize the boundary of the work area, but it may contain some invalid areas outside the farm plot due to the algorithm characteristics, and direct use will cause deviation of the farm work range. The total MultiPolygon of the plot only limits the actual planting range of the farm, and needs to be combined with the convex hull candidate area to determine the effective area that is both in the candidate work range and belongs to the farm planting plot. Therefore, the intersection of the two needs to be calculated relying on the spatial overlay analysis algorithm, and the result needs to be stored in the spatial database to realize structured management. The specific implementation is as follows: First of all, the core logic and applicable type of spatial overlay analysis algorithm are clarified. Here, the vector overlay analysis algorithm based on topological relationship is adopted. This algorithm analyzes the topological relationship of two MultiPolygon, calculates and extracts the geometric area of the intersection part, and has the core advantage of accurately preserving the boundary details of the intersection area, which meets the requirements of spatial accuracy for farm work. Before executing the algorithm, topological preprocessing needs to be performed on the two MultiPolygon objects. Topological preprocessing refers to repairing the topological errors that may exist in the MultiPolygon, so as to avoid failure of the overlay operation due to errors. The topological checking function of the geographic data processing library is called. If it is detected that the convex hull MultiPolygon has self-intersecting edges (some edges intersect with other edges of the same object), the repair function is automatically called to eliminate the error. If the total MultiPolygon of the plot has a non-closed boundary problem, the first and last vertex coordinates are inconsistent, then the first vertex coordinate is automatically supplemented to the end of the sequence to ensure that both MultiPolygon meet the operation conditions of no topological error, closed boundary, and no repeated vertex. After preprocessing, the spatial overlay analysis algorithm is started to calculate the intersection area. The specific process is as follows: First, the outer rectangle is quickly screened to calculate the outer rectangle of the two MultiPolygon. If the outer rectangle has no intersection, it is directly determined that the two MultiPolygon are disjoint and have no intersection area. If the outer rectangle has intersection, the attribution relationship of the MultiPolygon vertex is further determined by the ray method, and the existence of the intersection relationship between the two is preliminarily determined. Traverse all the edges of the two polygons MultiPolygon, the edges of the convex hull MultiPolygon and the edges of the total plot MultiPolygon, and perform intersection detection on each pair of edges. Whether the two edges intersect is determined by the vector cross product. If they intersect, the longitude and latitude coordinates of the intersection point are solved by a linear equation system, all intersection points are recorded and de-duplicated to avoid repeated calculation of the same intersection point, for example, the edge AB (coordinates A1, A2) of the convex hull MultiPolygon intersects the edge CD (coordinates C1, C2) of the total plot MultiPolygon, and the intersection point P is calculated and added to the intersection point set. All intersection points and original vertices located inside both MultiPolygon are integrated into a new point set, which is rearranged in counterclockwise order to ensure that the direction of the MultiPolygon outer boundary is correct, and then the point set is checked to see if it forms a closed contour. If the first and last vertices are not consistent, the first vertex coordinate is supplemented to the end, and finally a closed longitude and latitude coordinate sequence is generated. The MultiPolygon formed by this sequence is the intersection area of the convex hull MultiPolygon and the total plot MultiPolygon, which is the final operation area boundary data.

[0038] After generating the operation area boundary data, it needs to be verified for effectiveness to ensure that it meets the needs of agricultural machinery operation. The verification content includes boundary integrity, spatial rationality and coordinate accuracy, for example, the area of a certain operation area boundary data is only 80 square meters (less than the preset area of 100 square meters of a small farm), which is determined to be invalid and needs to be checked again. If the coordinates exceed the latitude and longitude interval of the farm, the coordinates are abnormal and need to be corrected by tracing back the coordinate conversion step. Only the operation area boundary data that passes all the verifications will enter the subsequent storage process, and then the operation area boundary data will be stored in the spatial database. The spatial database refers to a database system that supports the storage and spatial query of geographic spatial data. The storage core is the structured management of geographic data through the geometry type field. The geometry type field is a field type in the spatial database specially used to store geographic geometric objects. Here, the Geometry type is selected, and two key parameters need to be specified, namely the coordinate system and the geometric shape. The specific storage process is as follows: A job area management table is created, including the following core fields: job area ID (primary key, uniquely identifying each record), farm ID (associated with the farm information table, identifying the farm to which it belongs), plot ID (associated with the plot information table, identifying the plot involved), geom (geometry type field, storing job area boundary data), generation time (timestamp of record data generation), and threshold parameter (target index threshold when the job area is generated, facilitating subsequent tracing). The definition statement for the geom field is geomGeometry(MultiPOLYGON, 4326), which explicitly stores MultiPolygon and uses the WGS84 coordinate system. The longitude and latitude coordinate sequence of the job area boundary data is converted into a geometry data format supported by the spatial database, which uses the Well-Known Text (WKT) format, a text format that can visually describe geometric objects. The converted Geometry object is written into the job area management table along with other field values through an SQL insert statement.

[0039] Through the above process, the spatial accuracy of the job area boundary data is ensured, and the spatial database is used to manage and efficiently query the data for a long time. The subsequent closed-loop execution unit reads the data to correct it for the management personnel, and issues it to the intelligent agricultural machinery terminal to perform the job, laying a structured data foundation. At the same time, it is convenient for subsequent tracing of the distribution of job areas in different periods, assisting the farm in reviewing and optimizing the job effect.

[0040] The second object of the present application is to provide a system for implementing any of the above-mentioned satellite remote sensing driven intelligent job area generation management methods, comprising: The data preprocessing unit 1 is used to obtain the farm plot boundary vector data and calculate the circumscribed rectangle of the union of all plot boundaries to determine the geographic range of satellite remote sensing data, download Sentinel 2 satellite remote sensing data and perform resampling and cloud removal processing, calculate the crop vigor index and generate an image file according to the index, and then use the farm plot boundary vector data to crop the image file to generate a cropped image file; The intelligent analysis unit 2 is used to adaptively select the target index based on the current growth stage of the crop and set the threshold, read the cropped image file to extract the pixel matrix, generate a binary image based on the threshold, and perform an opening operation on the binary image to eliminate burrs and noise, traverse the binary image to perform connected component analysis to delete connected components with an area smaller than a predetermined area, and calculate the convex hull vertex pixel coordinates of the remaining connected components; The space optimization unit 3 is used for converting the convex hull vertex pixel coordinates into latitude and longitude coordinates and combining into a MultiPolygon object, combining all the plot boundaries of the farm into a MultiPolygon object, calculating the intersection of the two MultiPolygon objects to generate the job area boundary data and storing into the spatial database; The closed loop execution unit 4 is used for reading the job area boundary data from the spatial database for correction by the management personnel, and after correction, the final job range is issued to the intelligent agricultural machine terminal through a wireless communication protocol, the agricultural operation is executed, and the closed loop management process is completed.

[0041] The application obtains the plot boundary vector data of the farm through the geographic information system, calculates the circumscribed rectangle to determine the remote sensing range, and then downloads the Sentinel 2 satellite data; calculates the crop multiple vigor indexes through resampling and cloud removal processing, generates the image file in the plot according to the crop vigor index; and adaptively selects the target index and threshold according to the crop growth stage, generates a binary image and performs an open operation to remove noise; finally, the small area region is deleted through connected domain analysis, the convex hull vertex of the remaining region is calculated and converted into latitude and longitude, and the intersection with the plot boundary is calculated to generate the job area boundary data; after correction, the intelligent agricultural machine terminal is issued, the closed loop management is completed, the accurate generation of the job area and the efficient management are realized, and the modern agricultural precision operation demand is adapted.

[0042] The basic principles, main features and advantages of the application are shown and described. It should be understood by those skilled in the art that the application is not limited by the above examples, the above examples and descriptions in the specification are only preferred examples of the application, and are not used to limit the application, various changes and improvements of the application can be made without departing from the spirit and scope of the application, and these changes and improvements all fall within the scope of the claimed application. The scope of protection of the application is defined by the appended claims and their equivalents.

Claims

1. A satellite remote sensing-driven intelligent generation and management method for agricultural operation areas, characterized in that: Includes the following steps: S1. Obtain farm plot boundary vector data and calculate the bounding rectangle of the set of all plot boundaries to determine the geographic range of the satellite remote sensing data; S2. Download Sentinel-2 satellite remote sensing data, perform band fusion and cloud removal processing to calculate crop growth index, generate image files based on crop growth index, and then use farm plot boundary vector data to crop the image files to generate cropped image files. S3. Select the target index and set the threshold based on the current growth stage of the crop, read the cropped image file, extract the pixel matrix, and combine the pixel matrix with the set threshold to generate a binary image. When the pixel value is less than the threshold, it is set to 1 to indicate the area to be worked on, otherwise it is set to 0. Perform an opening operation on the binary image to eliminate burrs and noise. S4. Traverse the binary image and perform connected component analysis. Delete connected components with an area smaller than the preset area. Calculate the pixel coordinates of the convex hull vertices for the remaining connected components and convert the pixel coordinates of the convex hull vertices to latitude and longitude coordinates. During the conversion, use the metadata in the image file, including the latitude and longitude coordinates of the top left corner and the resolution parameter. Combine the latitude and longitude coordinates of the convex hull vertices into a MultiPolygon object. At the same time, combine the boundaries of all plots in the farm into a MultiPolygon object. Calculate the intersection of the two MultiPolygon objects to generate the boundary data of the work area and store it in the spatial database. S5. Finally, read the boundary data of the work area from the spatial database and allow corrections based on the actual situation. After correction, the final work area is sent to the intelligent agricultural machinery terminal to execute agricultural operations and complete the closed-loop management process.

2. The intelligent generation and management method for agricultural operation areas driven by satellite remote sensing according to claim 1, characterized in that: The acquisition of farm plot boundary vector data is specifically achieved by importing pre-stored farm vector data into a geographic information system. The pre-stored farm vector data contains a sequence of individual plot boundary coordinates, and the circumscribed rectangle of the plot boundary set is calculated.

3. The intelligent generation and management method for agricultural operation areas driven by satellite remote sensing according to claim 2, characterized in that: The downloaded Sentinel-2 satellite remote sensing data is automatically acquired based on a public data interface, and the data of different resolution bands are unified to the same resolution through resampling. A cloud coverage mask is generated using CLDPRB band data and a preset threshold, and the values ​​of other bands covered by clouds are set to zero. When calculating crop growth index, multiple crop growth indices are generated according to the preset index calculation formula and directly output as image files. Then, the image files are cropped using farm plot boundary vector data. Through vector and raster overlay analysis, only pixel areas within the plots are retained to generate cropped image files.

4. The intelligent generation and management method for agricultural operation areas driven by satellite remote sensing according to claim 1, characterized in that: The selection of target indices and setting thresholds based on the current growth stage of crops specifically includes: The growth stages are divided into emergence, growth, and maturity stages by dynamic matching of the growth model. During the emergence stage, an index type sensitive to seedling density is selected; during the growth stage, an index type sensitive to canopy cover is selected; and during the maturity stage, an index type sensitive to nutrient status is selected. The threshold setting is adaptively generated based on the classification model trained on historical data. After reading the cropped image file, the index value matrix is ​​extracted.

5. The intelligent generation and management method for agricultural operation areas driven by satellite remote sensing according to claim 4, characterized in that: When generating a binary image by combining the pixel matrix and a set threshold, the exponent value is compared with the threshold pixel by pixel. When the exponent value is less than the threshold, it is marked as a region that needs to be processed and assigned a value of 1. Otherwise, it is marked as a non-processing region and assigned a value of 0. The opening operation on the binary image is performed using morphological filtering, which first erodes and then dilates to eliminate burrs and noise.

6. The intelligent generation and management method for agricultural operation areas driven by satellite remote sensing according to claim 5, characterized in that: When performing connected component analysis on a binary image, a region growing algorithm is used to identify adjacent pixel regions and calculate the actual area of ​​each connected component. Connected components with areas smaller than a preset area are deleted. The preset area is dynamically adjusted according to the task requirements. The pixel coordinates of the convex hull vertices are calculated for each of the remaining connected components. The Andrew algorithm is used to solve for the convex hull vertices with the pixel coordinates of the boundary points of the connected components as input.

7. The intelligent generation and management method for agricultural operation areas driven by satellite remote sensing according to claim 6, characterized in that: When converting the pixel coordinates of the convex hull vertex to latitude and longitude coordinates, the metadata in the image file is directly read, including the latitude and longitude coordinates of the top left corner and the resolution parameters. The conversion process achieves the mapping from pixel coordinates to geographic coordinates through coordinate affine transformation.

8. The intelligent generation and management method for agricultural operation areas driven by satellite remote sensing according to claim 7, characterized in that: When combining the latitude and longitude coordinates of the convex hull vertices [1] into a MultiPolygon object, the latitude and longitude coordinate sequence after the transformation of a single convex hull is first encapsulated into a single Polygon object in counterclockwise order, and then multiple Polygon objects are combined into a MultiPolygon object. When combining the farm plot boundaries into a MultiPolygon object, the boundaries of a single plot are first encapsulated into a single Polygon object, and then the Polygon objects of all farm plots are combined into a MultiPolygon object.

9. The intelligent generation and management method for agricultural operation areas driven by satellite remote sensing according to claim 8, characterized in that: When calculating the intersection of two MultiPolygon objects, a spatial overlay analysis algorithm is used to find the intersection area between the convex hull MultiPolygon object and the plot boundary MultiPolygon object, generating the work area boundary data, and storing this data in a spatial database using a geometric type field.

10. A system for implementing the satellite remote sensing-driven intelligent generation and management method for agricultural operation areas as described in any one of claims 1-9, characterized in that, include: The data preprocessing unit (1) is used to acquire farm plot boundary vector data and calculate the bounding rectangle of the set of all plot boundaries to determine the geographical range of satellite remote sensing data. It downloads Sentinel 2 satellite remote sensing data and performs band resampling and cloud removal. It calculates the crop growth index and generates image files based on the index. Then, it uses the farm plot boundary vector data to crop the image files and generate cropped image files. The intelligent analysis unit (2) is used to adaptively select the target index and set the threshold based on the current growth stage of the crop, read the cropped image file to extract the pixel matrix, combine the threshold to generate a binary image and perform an opening operation on the binary image to eliminate burrs and noise, traverse the binary image to perform connected component analysis to delete connected components with an area smaller than the preset area, and calculate the convex hull vertex pixel coordinates of the remaining connected components. The spatial optimization unit (3) is used to convert the pixel coordinates of the convex hull vertex into latitude and longitude coordinates and combine them into a MultiPolygon object. At the same time, it combines the boundaries of all plots in the farm into a MultiPolygon object, calculates the intersection of two MultiPolygon objects to generate the boundary data of the work area and stores it in the spatial database. The closed-loop execution unit (4) is used to read the boundary data of the work area from the spatial database for the management personnel to correct. After correction, the final work range is sent to the intelligent agricultural machinery terminal through the wireless communication protocol to perform agricultural operations and complete the closed-loop management process.

Images