Satellite data-based farmland planning analysis method and system

Abstract

The application discloses a farmland planning analysis method and system based on satellite data, relates to the technical field of farmland planning analysis, and comprises the following steps: obtaining multi-satellite image data of a target farmland area, pre-processing and performing semantic segmentation, constructing a binary image, fitting line segment endpoints, forming a farmland closed polygon, obtaining agricultural machinery parameters, forming constraint conditions, performing regional decomposition on the farmland closed polygon based on the constraint conditions, forming a plurality of sub-regions with continuous operation attributes, constructing a bidirectional optional coverage path and a state set, judging the direction consistency of the path connection relationship between the sub-regions according to the coverage path, generating a complete path for division and evaluation based on the judgment result, and performing local path reconstruction according to the evaluation result to obtain a complete continuous path as the farmland planning analysis result. The application significantly improves the intelligent level of farmland planning analysis.

Description

Technical Field

[0001] This invention relates to the field of farmland planning and analysis technology, and in particular to a farmland planning and analysis method and system based on satellite data. Background Technology

[0002] With the rapid development of remote sensing technology and agricultural informatization, farmland planning and operation path analysis technologies based on satellite data have gradually become an important component of precision agriculture. Existing technologies typically identify and extract boundaries of farmland areas using multi-source satellite imagery (such as Sentinel and Landsat), and then combine this with geographic information systems (GIS) to spatially model the farmland. Based on this, various coverage path planning methods have been proposed to address the problem of agricultural machinery operation path planning, such as path planning methods based on regular grids, reciprocating coverage methods based on parallel scan lines, and path optimization methods based on graph search. These methods can, to a certain extent, achieve automated coverage operations in farmland areas.

[0003] However, existing technologies tend to generate many invalid turning paths when dealing with irregular farmland boundaries, leading to decreased operational efficiency. Furthermore, they are often based on simple topological partitioning or regular grid partitioning, making it difficult to guarantee the continuity of the internal structure of sub-regions, which in turn affects the feasibility and coherence of the paths. In addition, they ignore the minimum turning radius constraints of agricultural machinery and the issue of path direction consistency, resulting in the generated connecting paths having infeasible turns or sharp turns in actual execution. Secondly, the lack of a quantitative evaluation mechanism for path quality and the failure to effectively control the degree of path detours may result in a large redundant length of the overall path, thereby increasing operation time and energy consumption. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a farmland planning and analysis method and system based on satellite data, which solves the problems of existing technologies that make it difficult to ensure the continuity of the internal structure of sub-regions, the generated connection paths having infeasible turns or sharp turns in actual implementation, and the inability to effectively control the degree of path detours.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a farmland planning and analysis method based on satellite data, comprising, Multi-satellite imagery data of the target farmland area is acquired, preprocessed, and semantically segmented. A binarized map is constructed and the endpoints of line segments are fitted to form a closed polygon of farmland. Obtain agricultural machinery parameters and formulate constraints; Based on constraints, the closed polygon of farmland is decomposed into multiple sub-regions with continuous operation attributes. A set of bidirectional selectable coverage paths and states is constructed, and the path connection relationship between sub-regions is constructed for the coverage path to determine the directional consistency. Based on the judgment results, a directed weighted graph is constructed to determine the access order and coverage direction of the sub-regions. Then, a complete path is generated for partitioning and evaluation. Based on the evaluation results, a local path reconstruction is performed to obtain a complete continuous path as the farmland planning analysis result.

[0007] As a preferred embodiment of the farmland planning and analysis method based on satellite data described in this invention, the preprocessing and semantic segmentation of multi-satellite image data of the target farmland area are performed as follows: Multi-satellite imagery data was acquired through a GIS platform, and coordinate transformation and resampling were performed. The imagery data was then normalized and overlaid to obtain feature vectors. The multi-satellite imagery data included Sentinel-2 multispectral imagery, Landsat-8 imagery, and Gaofen series GF satellite imagery. Based on the feature vector, the probability of farmland category is obtained; farmland category includes farmland, water body, uncultivated land, and shrubland.

[0008] As a preferred embodiment of the farmland planning and analysis method based on satellite data described in this invention, the step of constructing a binarized map and fitting the endpoints of line segments to form a closed polygon of farmland is as follows: Based on probability comparison, the binarized values ​​required for the binarized image are generated through a preset judgment threshold. The binarized image is then constructed, and contours are found to form a continuous set of contour points. Based on a continuous set of contour points, the vertical distance is calculated, and then a set of polygon vertices is generated and connected using a preset deviation threshold to form a closed polygon.

[0009] As a preferred embodiment of the farmland planning and analysis method based on satellite data described in this invention, the step of obtaining agricultural machinery parameters and forming constraints involves finding the minimum turning radius and operating width of the agricultural machinery in the agricultural machinery design book, using the minimum turning radius as the basic parameter of curvature constraint in path planning, generating constraint formulas, and defining safe operating distances based on the operating width of the agricultural machinery and actual farmland operation requirements.

[0010] As a preferred embodiment of the farmland planning and analysis method based on satellite data described in this invention, the step of performing region decomposition on the closed polygon of farmland is as follows: Determine the scanning direction; Scan lines are generated along the scanning direction to perform intersection detection. If they intersect, the intersection coordinates are calculated, sorted, and paired to form several intervals. Otherwise, the intersection coordinates are not calculated, sorted, or matched. Based on several intervals, perform interval segmentation to generate sub-regions.

[0011] As a preferred embodiment of the farmland planning and analysis method based on satellite data described in this invention, the construction of a bidirectional selectable coverage path and state set is specifically as follows: Within the sub-region, coverage path segments are generated again according to the scanning direction and connected back and forth to form a bidirectional optional path for continuous path judgment. Extract the positions of the start and end points and their corresponding path directions from the bidirectional optional paths, and construct a state set, including a forward path state set and a reverse path state set.

[0012] As a preferred embodiment of the farmland planning and analysis method based on satellite data described in this invention, the step of determining directional consistency based on the path connection relationships between sub-regions constructed according to the coverage path is as follows: Based on the state set, the path endpoint state of the previous sub-region and the path start state of the next sub-region are obtained as candidate connection state pairs, and connection paths are generated under the constraints of the constraints. After enumerating candidate paths and selecting the optimal connection path as the edge weight, a connection relationship structure is generated, and optional and non-optional markings are performed based on a preset consistency threshold.

[0013] As a preferred embodiment of the farmland planning and analysis method based on satellite data described in this invention, after constructing a directed weighted graph to determine the access order and coverage direction of sub-regions, a complete path is generated for partitioning and evaluation, as follows: Based on bidirectional optional paths, independent nodes are generated and marked as optional or non-optional to form an edge set. Based on the node and edge set, a directed weighted graph is formed. Select a node as the starting node for access, generate an initial path sequence, perform edge swapping and expansion, and obtain the complete path; The path quality value is obtained by dividing the path based on the complete path, calculating the actual path length and the straight-line distance, and then performing a ratio calculation.

[0014] As a preferred embodiment of the farmland planning and analysis method based on satellite data described in this invention, the step of performing local path reconstruction based on the evaluation results to obtain a complete and continuous path involves setting a quality threshold. If the path quality value is greater than the quality threshold, the path segment is determined to be a low-quality path segment; otherwise, the path segment is determined to be a high-quality path segment. Based on low-quality path segments, the access order is swapped and the path direction is switched within the set of sub-regions involved in the path segment. The total length of the connecting path is recalculated, and the scheme with the shortest total length of the connecting path is selected for replacement. This process continues until all path segments meet the quality threshold. By sequentially splicing sub-region coverage paths, connecting paths, and sub-region coverage paths, all sub-region paths are connected into a complete continuous path, and then the complete continuous path is used as the result of farmland planning analysis.

[0015] Secondly, this invention provides a farmland planning and analysis system based on satellite data, comprising: The farmland closed polygon generation module is used to acquire multi-satellite image data of the target farmland area, perform preprocessing and semantic segmentation, construct a binarized map and fit the endpoints of line segments to form farmland closed polygons. The agricultural machinery parameter acquisition module is used to acquire agricultural machinery parameters and form constraints. The path generation and constraint module is used to perform region decomposition on closed polygons of farmland based on constraints, forming multiple sub-regions with continuous operation attributes, constructing bidirectional selectable coverage paths and state sets, and determining the directional consistency of path connection relationships between sub-regions for coverage paths. The evaluation and reconstruction module is used to construct a directed weighted graph based on the judgment results to determine the access order and coverage direction of the sub-regions, generate a complete path for division and evaluation, and then perform local path reconstruction based on the evaluation results to obtain a complete continuous path as the farmland planning analysis result.

[0016] The beneficial effects of this invention are as follows: By introducing an adaptive scanning direction determination mechanism based on covariance feature analysis, the main extension direction can be automatically extracted according to the shape of the farmland boundary, thereby effectively reducing the number of turns in the path generation process, improving the overall continuity of the path and operational efficiency. Furthermore, by using a region decomposition method based on the variation of scan line intervals, complex farmland is divided into structurally consistent sub-regions, ensuring that the paths within each sub-region have good continuity and executability. In addition, a curvature constraint mechanism is introduced during the sub-region connection process, and the connection path is constructed in conjunction with the minimum turning radius of agricultural machinery, ensuring that the generated path is executable in actual agricultural machinery operations. Moreover, a ratio evaluation mechanism between path length and straight-line distance is constructed to quantitatively analyze the path quality, and local path reconstruction is performed based on the evaluation results, thereby effectively reducing the degree of path detour and optimizing the overall path length. Therefore, this invention achieves significant improvements in path continuity, executability, and path quality control, and significantly enhances the intelligence level of farmland planning and analysis. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of the farmland planning and analysis method based on satellite data in Example 1.

[0019] Figure 2 This is a structural diagram of the farmland planning and analysis system based on satellite data in Example 1.

[0020] Figure 3 This is a flowchart of the local path reconstruction process in Example 1.

[0021] Figure 4 This is a flowchart of constructing a directed weighted graph in Example 1. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0025] Example 1, referring to Figures 1-4 This is the first embodiment of the present invention, which provides a farmland planning and analysis method based on satellite data, including the following steps: S1. Acquire multi-satellite image data of the target farmland area, perform preprocessing and semantic segmentation, construct a binarized map and fit the endpoints of line segments to form a closed polygon of farmland; S1.1. Obtain multi-satellite imagery data through a GIS platform, perform coordinate transformation and resampling, normalize and overlay the imagery data to obtain feature vectors; the multi-satellite imagery data includes Sentinel-2 multispectral imagery, Landsat-8 imagery, and Gaofen series GF satellite imagery. Specifically, Sentinel-2 multispectral imagery, Landsat-8 imagery, and GF satellite imagery of the target farmland area are acquired from the GIS platform to form multi-satellite imagery data; the pixel coordinates of the multi-satellite imagery data are converted into geographic coordinates through projection transformation; subsequently, bilinear interpolation is used to resample the different satellite imagery data to form imagery data with uniform spatial resolution. The pixel values ​​in the image data of different bands are normalized by the min-max normalization method, and the pixel values ​​in the normalized multi-band image data are superimposed to form the feature vector of each pixel. It should be noted that multi-satellite imagery data includes multiple band images, and the number and range of bands corresponding to different satellite images are different; while pixel value overlay refers to vector concatenation according to the band dimension rather than numerical addition.

[0026] S1.2. Based on the feature vector, obtain the probability of farmland category; farmland categories include farmland, water body, uncultivated land, and shrubland; Specifically, based on the feature vector of each pixel, semantic segmentation is performed, and the probability that the pixel belongs to the farmland category is output; the farmland category includes farmland, water bodies, uncultivated land, and shrubland; It should be noted that semantic segmentation requires model building. In terms of model building, a convolutional neural network based on an encoder-decoder structure is used as the semantic segmentation model. The encoder part is used to extract multi-scale spatial features, and the decoder part is used to restore spatial resolution and output pixel-by-pixel classification results. For example, in one implementation, a U-Net network structure is adopted, in which the encoder part uses a pre-trained ResNet-50 model to improve feature extraction capability and accelerate model convergence speed; the decoder part achieves fine segmentation of farmland boundaries by upsampling layer by layer and combining skip connections to fuse low-level spatial information and high-level semantic information; during the training phase, supervised learning training is performed using a remote sensing dataset with pixel-level annotations, and the cross-entropy loss function is used; during the training process, the Adam optimizer is used to update the model parameters, and during the iteration process, when the loss value of the loss function no longer decreases, the iteration stops and the trained model is output.

[0027] S1.3. Based on probability comparison, after generating the required binarization values ​​for the binarization map through a preset judgment threshold, the binarization map is constructed, and contours are found to form a continuous set of contour points. Specifically, a judgment threshold is set, the probability is compared with the judgment threshold, a binary value of each pixel is generated, a binary image is constructed, and contours are found in the binary image to form a continuous set of contour points. It should be noted that when comparing probability with the judgment threshold, if the judgment threshold is too low, non-farmland areas are easily misclassified as farmland, leading to an increased false detection rate; if the judgment threshold is too high, edge farmland areas are easily missed, resulting in incomplete farmland boundaries. Therefore, the judgment threshold is set to an example range of 0.4 to 0.7; it can be fine-tuned later according to the crop type and image quality of different regions. Secondly, finding contours in a binarized image can be done using existing techniques; For example, in one implementation, the cv2.findContours() function in OpenCV is used to extract the contours of connected regions.

[0028] S1.4 Based on a continuous set of contour points, after calculating the vertical distance, a set of polygon vertices is generated and connected to form a closed polygon using a preset deviation threshold. Specifically, the points in the contour point set are arranged in the order of contour output, and the first and last points of the contour point sequence are used as the two endpoints of the initial fitted line segment. For all points between the first and last points, calculate the perpendicular distance from each point to the first and last points; Sort the vertical distances in descending order and find the maximum vertical distance; Set a deviation threshold. If the maximum vertical distance is greater than the deviation threshold, the point corresponding to the maximum vertical distance is used as the dividing point. The initial fitted line segment is divided into segments to form the first subsequence and the second subsequence. Otherwise, the endpoints are used as vertices of the fitted polygon. The first and second subsequences are recursively divided until the vertical distance of all points does not exceed the deviation threshold. Then, all endpoints are used as vertices of the fitted polygon to generate a set of vertices of the fitted polygon. Arrange the vertices in the fitted polygon vertex set according to the original scan order of the contour to obtain the polygon vertex set; then connect the vertices in the order of endpoint generation, and then connect the first and last points to form a closed polygon. The probability of farmland category is expressed as: In the formula, Represents pixels The probability of farmland category This represents the softmax activation function. Represents pixels eigenvectors, Represents the weight matrix. Indicates the bias term; Generate the binary value for each pixel using the following expression: In the formula, Represents pixels The binarized value, Represents pixels It belongs to farmland. Represents pixels It is not farmland. Indicates the threshold for judgment; The vertical distance from each point to the first and last points is calculated using the following expression: In the formula, Indicates the first The perpendicular distance from each point to the first and last points Indicates the coordinates of the first point. Represents the coordinates of the tail point. Indicates the first The coordinates of the points This represents the absolute value operation; It should be noted that: the closed polygon defines the spatial range of farmland operations; and the deviation threshold is set to ensure improved planning accuracy; if the deviation threshold is too small, the fitted polygon will contain a large number of redundant vertices, which will significantly increase the complexity of subsequent path planning calculations; if the deviation threshold is too large, the polygon boundary will deviate significantly from the actual farmland boundary, reducing planning accuracy. For example, in one implementation, the deviation threshold is set to a range of 0.5 meters to 5 meters; and in order to achieve a good balance between accuracy and computational efficiency and to adapt to the spatial accuracy of 10-meter resolution remote sensing images, 1 meter can be used as the initial value of the deviation threshold.

[0029] S2. Obtain agricultural machinery parameters and form constraints; Specifically, find the minimum turning radius and working width of the agricultural machinery in the agricultural machinery design book; The minimum turning radius is used as the basic parameter for curvature constraints in path planning to generate constraint formulas. Define safe operating distances based on the operating width of agricultural machinery and actual farmland operation needs; The continuous curve is expressed as: In the formula, Represents a continuous curve. Indicates path parameters (arc length); The constraint formula is expressed as follows: In the formula, Indicates the constraint curvature. Indicates the minimum turning radius of agricultural machinery; Define a safe working distance using the following expression: In the formula, Indicates safe working distance. Indicates the working width of agricultural machinery. Indicates safety redundancy; It should be noted that: safety redundancy It is set based on GNSS error, control error, and terrain error; For example, in one implementation, the errors are: GNSS error: 0.02~0.1 m; control error: 0.05~0.2 m; terrain error: 0.05~0.3 m; By summing the various errors, a safety redundancy is achieved. The range of values ​​is To ensure coverage of all error sources, 0.3m can be used as the default value for example. It should also be noted that the defined safe working distance ensures that adjacent paths maintain a distance no less than the safe working distance, in order to avoid overlapping or missing work areas, thereby improving the efficiency and quality of farmland operations.

[0030] S3. Based on the constraints, perform region decomposition on the closed polygon of farmland to form multiple sub-regions with continuous operation attributes, construct bidirectional selectable coverage paths and state sets, and determine the directional consistency of the path connection relationship between sub-regions for the coverage path. S3.1 Determine the scanning direction; Specifically, the geometric center of the farmland area is calculated based on the closed polygon. After obtaining the center point, the boundary points are decentralized to obtain decentralized coordinates; Based on decentralized coordinates, the covariance matrix is ​​constructed by mean normalizing the coordinates of boundary points. Then, the corresponding eigenvectors and eigenvalues ​​are obtained by eigenvalue decomposition of the covariance matrix. Based on the corresponding eigenvectors and eigenvalues, the eigenvector corresponding to the largest eigenvalue is selected to calculate the scanning direction angle. By using a two-dimensional coordinate rotation transformation method, the farmland area in the original coordinate system is mapped to the new coordinate system through rotation transformation, so that the scanning direction is aligned with the horizontal axis of the new coordinate system; The geometric center calculation is expressed as follows: In the formula, Indicates the center point coordinates This represents the total number of boundary points. Indicates the first The coordinates of the boundary points; The angle of the scanning direction is calculated using the following expression: In the formula, Indicates the scanning direction angle. Representing the eigenvector The directional component in (i.e.) Axis coordinates and (axis coordinates) Represents the arctangent function; It should be noted that: Centralization involves subtracting the geometric center coordinates from the original coordinates of each boundary point, thus redistributing all boundary points relative to the center point and eliminating the impact of translation on subsequent direction calculations. Next, mean normalization calculates the mean and standard deviation of all coordinates after centralization and standardizes each coordinate component, avoiding deviations in covariance calculations due to scale differences in different directions. The covariance matrix is ​​constructed by organizing the coordinate data of all boundary points into a two-dimensional dataset and statistically analyzing the correlation between the horizontal and vertical coordinates, forming a covariance matrix that describes the spatial distribution directionality of the point set. Furthermore, the eigenvector corresponding to the largest eigenvalue is selected to calculate the scanning direction angle because this direction represents the most significant extension of the farmland area. Determining the scanning direction in this way effectively reduces the number of turns in the subsequent path and improves path continuity.

[0031] S3.2 Generate scan lines along the scanning direction and perform intersection detection. If they intersect, calculate the intersection coordinates, sort them, and pair them up to form several intervals. Otherwise, do not calculate the intersection coordinates, sort them, or match them. Specifically, the agricultural machinery operating width is set as the scanning interval; then, along the scanning direction angle, scanning lines are generated starting from the minimum projection position of the safe operating interval in the scanning direction angle, thus obtaining a set of scanning lines; By traversing each edge of the closed polygon, we check whether the edge intersects with the scan line. If they intersect, we calculate the intersection point between the scan line and the edge. Coordinates that do not intersect will not be included in the intersection point. Coordinate calculation; According to the intersection point The coordinates are sorted and paired in order to form several intervals; The expression for calculating the intersection of the scan line and the edge is: In the formula, Indicates the x-coordinate of the intersection point. Indicates the starting coordinates of the edge. Represents the coordinates of the endpoint of the edge. Indicates the height of the scan line; It should be noted that when determining intersection, And the edge is a horizontal edge. If the line passes through a vertex, it will not be included in the intersection calculation; if the scan line happens to pass through a vertex, the intersection will only be recorded once. Moreover, the intervals are constructed for all intersections obtained from the same scan line, and each interval represents the effective coverage segment of the scan line within the farmland area. For example, in one implementation, all intersections are arranged according to... The coordinates are sorted from smallest to largest. According to the sorting result, the intersection points are paired up to form several intervals. For example, the first intersection point and the second intersection point form the first interval, the third intersection point and the fourth intersection point form the second interval, and so on. If the number of intersection points is odd, the last intersection point is ignored.

[0032] S3.3. Perform interval segmentation based on several intervals to generate sub-regions; Specifically, the number of intervals in consecutive scan lines is compared. If the number of intervals in adjacent scan lines changes, it indicates that there is a change in the topology at that location, and the location is marked as a key decomposition point. If the number of intervals in adjacent scan lines does not change, the intervals of adjacent scan lines are matched accordingly. For each pair of matched intervals, a strip region is generated, and the region between adjacent scan lines is filled into a continuous sub-region segment. Based on key decomposition points, the entire farmland area is divided into multiple sub-regions. After adding strip regions to the sub-regions, all sub-regions are organized to generate a set of sub-regions. It should be noted that during interval matching, to ensure the stability and uniqueness of the matching relationship, the intervals on each scan line need to be sorted according to their center positions. For example, the center position of an interval can be calculated by averaging the two endpoints of the interval. Subsequently, the sorted interval set is matched sequentially with the corresponding intervals on adjacent scan lines to form a one-to-one correspondence. After completing the interval matching, a strip region can be constructed for each pair of matched intervals. This strip region consists of four vertices, namely the two endpoints of the interval on the previous scan line and the two endpoints of the corresponding interval on the next scan line. Connecting them sequentially forms a quadrilateral region. Furthermore, this quadrilateral region represents a continuous coverage area segment between two scan lines. If no topological change is detected during the scan, the strip regions can be continuously accumulated and classified into the same sub-region. If a key decomposition point is detected, it indicates that the regional structure has changed. At this time, it is necessary to end the construction of the current sub-region and start creating a new sub-region from that position. In this way, the entire farmland area can be divided into multiple sub-regions with consistent topological structures. Each sub-region consists of multiple strip regions, and the internal structure is continuous without splitting or merging.

[0033] S3.4 Within the sub-region, generate coverage path segments again according to the scanning direction and connect them back and forth to form a bidirectional selectable path for path continuity judgment. Specifically, within each sub-region, coverage path segments are generated according to the same scanning direction; The covered path segments are connected in a back-and-forth manner to form a continuous path similar to a sawtooth shape; that is, odd-numbered paths are connected from left to right, and even-numbered paths are connected from right to left; then, a bidirectional optional path is constructed for each sub-region, including a forward path and a reverse path. A forward path is defined as connecting path segments in the order of generation, while a reverse path is connecting path segments in reverse order. For adjacent path segments, the connection distance is calculated using Euclidean distance; a distance threshold is set, and if the connection distance is less than or equal to the distance threshold, the path is considered continuous; otherwise, the sub-region is decomposed again. When the number of intervals between adjacent scan lines changes, the expression is: In the formula, Indicates the first The number of intervals within a scan line. Indicates the first scan lines, Indicates the first The number of intervals within a scan line. Indicates the first One scan line; If the number of intervals between adjacent scan lines remains unchanged, the expression is: The connection distance is calculated using Euclidean distance, expressed as: In the formula, Indicates the first The path and the first The connection distance between the paths This represents Euclidean norm operations. Function representing the endpoint of a path segment , Indicates the first The first sub-region The coordinates of the endpoint of the path. Function representing the starting point of a path segment , Indicates the first The first sub-region The starting coordinates of the path; It should be noted that the distance threshold needs to be determined comprehensively based on the working width of the agricultural machinery, the path generation spacing, and the turning capability of the machinery. Since the scanning lines inside the sub-region in this invention are parallel lines generated according to the working width of the agricultural implement, the theoretical minimum spacing between adjacent path segments should be consistent with or close to the working width. That is, the lateral distance between adjacent path segments is mainly determined by the working width. For example, in one implementation, if the distance is too small, it indicates significant overlap between path segments or abnormal path generation, potentially leading to duplicate coverage and thus failing to meet reasonable connection conditions. Conversely, if the distance is too large, it indicates significant gaps between path segments, making it impossible to guarantee coverage continuity. In this case, the area needs to be re-decomposed or the scanning path regenerated. When the distance is close to the working width, it indicates that the spacing between path segments is consistent with the design spacing, meeting the standard coverage pattern, and the path connection is most reasonable. Agricultural machinery must be able to achieve path connection through continuous turns. To ensure the feasibility of the connected path, the distance threshold can be set to be less than or equal to twice the minimum turning radius. Therefore, the distance threshold should be between the working width and twice the minimum turning radius. The minimum value is taken from the middle value to generate the result.

[0034] S3.5 Extract the positions of the start and end points and the corresponding path directions in the bidirectional optional path, and construct a state set, including the forward path state set and the reverse path state set.

[0035] Specifically, for each bidirectional optional path generated in a sub-region, the positions of the starting point and the ending point of the path and the corresponding path direction are extracted to obtain two state sets for each sub-region: a forward path state set and a reverse path state set. Each state set contains the starting point position, the ending point position, and the corresponding path direction information. The starting position, ending position, and path direction are uniformly represented as state vectors; this process is repeated to generate the state set of all sub-regions. It should be noted that for each sub-region, its forward path corresponds to a forward state vector, and its reverse path corresponds to a reverse state vector, so that each sub-region corresponds to two state vectors; in this way, the path information can be converted from a geometric path form into a structured data form.

[0036] S3.6 Based on the state set, obtain the path endpoint state of the previous sub-region and the path start state of the next sub-region as candidate connection state pairs, and generate connection paths under the constraints of the constraints. Specifically, for any two sub-regions, the path endpoint state of the preceding sub-region and the path start state of the following sub-region are selected as candidate connection state pairs. A path construction method based on the combination of circular arcs and straight lines is adopted to construct a connecting path between the path endpoint and the path start point that satisfies curvature constraints and consists of circular arc segments and straight line segments; When constructing the connection path, the length of the arcs in the connection path is calculated, and the length of the straight line segment is determined by the Euclidean distance between the tangent points of the two arcs; the arc lengths and straight line segment lengths are summed to generate the total length of the connection path; The expression for calculating the arc length in the connecting path is: In the formula, Indicates the length of the first arc. Indicates the turning angle; The expression for determining the length of a line segment is: In the formula, Indicates the length of a straight line segment. , These represent the coordinates of the first and second arc tangent points, respectively. The total length of the connection path is expressed as: In the formula, Indicates the total length of the connection path. Indicates the length of the second arc; It should be noted that since each sub-region contains forward path states and reverse path states, in actual operation, it is necessary to traverse all possible state combinations, that is, to construct corresponding connection candidate state pairs by combining the forward end state of the previous sub-region with the forward start state of the next sub-region, the forward end state of the previous sub-region with the reverse start state of the next sub-region, the reverse end state of the previous sub-region with the forward start state of the next sub-region, and the reverse end state of the previous sub-region with the reverse start state of the next sub-region, thereby forming a complete set of connection candidates. The connecting path consists of three parts: the first segment is the first arc segment from the end of the path into the turn along the current direction; the second segment is the straight transition segment connecting the two arc segments; and the third segment is the second arc segment transitioning from the straight segment to the target direction. Furthermore, when constructing the first arc segment, it is necessary to determine an arc path with a minimum turning radius based on the spatial position and direction information of the path endpoint of the previous sub-region. The starting point of this arc is the path endpoint position, and its initial direction is consistent with the path endpoint direction. Subsequently, based on the geometric relationship between the target path direction and the current position, the turning direction (left turn or right turn) and the corresponding turning angle of this arc are determined, thus constructing the first arc segment. In this process, the turning angle is determined by the angle between the starting direction and the target transition direction. When constructing the second arc segment, a similar method is used. Based on the spatial position and direction information of the path starting point of the subsequent sub-region, the direction is gradually adjusted from the intermediate transition direction to the path starting direction to determine the corresponding turning direction and turning angle, and the second arc segment is constructed. The endpoint of this arc coincides with the path starting point position, and its ending direction is consistent with the path starting direction.

[0037] S3.7 After enumerating candidate paths and selecting the optimal connection path as the edge weight, generate the connection relationship structure and perform optional and non-optional marking through the preset consistency threshold; Specifically, after enumerating different combinations of turning directions and calculating the path length of the connection path generated for each enumeration form, the path with the smallest path length is selected as the optimal connection path between the corresponding sub-regions from all candidate paths that satisfy the curvature constraint. The optimal connection path is used as the edge weight, and then a connection relationship structure is formed with the path endpoint and path start point in all sub-regions. Based on the constructed connection structure, for any two points (end point, start point), the absolute difference between the direction angles of the two points is calculated to determine the consistency of direction. If the absolute difference is greater than the consistency threshold, the connection path structure is marked as unselectable; otherwise, it is marked as selectable. The expression for calculating the absolute difference between the direction angles of two points is: In the formula, This represents the absolute difference between the direction angles of two points. Indicates the previous subregion The direction of the destination, Indicates the previous subregion The starting direction; It should be noted that: Enumerating different combinations of turning directions, such as left turn-straight-left turn, right turn-straight-right turn, left turn-straight-right turn, etc.; for each combination of turning directions, firstly, the coordinates of the starting and ending points are transformed to unify them into a local coordinate system. Then, the corresponding path parameters are calculated based on the current turning combination, including the arc angle of the turning segment, the length of the straight segment, and the connection relationship between each segment. After obtaining the path parameters, the total length of the combined path is obtained by summing the lengths of each segment. During the calculation process, curvature constraint verification must be performed on each path segment to ensure that the turning radius is not less than the minimum turning radius of agricultural machinery, thereby ensuring that the path will not be infeasible in actual execution. Path combinations that do not meet the curvature constraints should be directly eliminated. Secondly, when using the optimal connection path as the edge weight to form a connection structure with the path endpoints and starting points of all sub-regions, it is necessary to extract the starting point, ending point, starting direction, and ending direction of the covered path for each sub-region. Then, the "covered path endpoint" of each sub-region is paired with the "covered path starting point" of other sub-regions to form candidate connection pairs. For each candidate connection pair, its corresponding optimal connection path is used as the connection weight, establishing the following data structure: Using the "sub-region endpoint" as the connection start node, the "sub-region start point" as the connection target node, and the optimal connection path length as the connection weight, these nodes and weight information are stored as a connection record. Then, all connection records are summarized to construct a complete connection relationship structure. It should also be noted that the consistency threshold is used to constrain the directional change between adjacent paths, ensuring that the path connection satisfies curvature constraints while avoiding excessive turning movements. However, due to the minimum turning radius constraint on agricultural machinery, when the directional change is too large, even if a path that meets the curvature constraint exists, the turning path length will increase significantly, thereby reducing overall operational efficiency. Therefore, the consistency threshold value range needs to be set based on the following two comprehensive factors: 1. Minimum turning radius of agricultural machinery: The larger the minimum turning radius, the weaker the steering ability of the machinery, and the smaller the allowable directional change range should be. Therefore, when the turning radius is large, a smaller consistency threshold should be selected to avoid producing large-angle turning paths. 2. Spatial distance between sub-regions: If the distance between two sub-regions is large, agricultural machinery will have enough space to make directional adjustments, and the consistency threshold should be appropriately relaxed in this case; conversely, when the distance between sub-regions is small, the consistency threshold should be tightened to avoid drastic turning within a short distance. Based on the combined factors of points one and two, the range of values ​​for the consistency threshold can be set as follows: The corresponding angle range is approximately to When the consistency threshold is less than If the directional constraints are too strict, a large number of potential feasible paths may be filtered out, affecting path connectivity and even causing path optimization to fail to converge; when the consistency threshold is greater than If the allowed directional changes are too large, it will introduce a large number of sharp turns, resulting in a significant increase in path length and a reduction in the efficiency of agricultural machinery operations; therefore, the consistency threshold can be set as follows: This allows agricultural machinery to complete a smooth turn in a single operation, meeting the steering capability constraints of common agricultural machinery.

[0038] S4. Based on the judgment results, after constructing a directed weighted graph to determine the access order and coverage direction of the sub-regions, a complete path is generated for division and evaluation. Then, based on the evaluation results, a local path reconstruction is performed to obtain a complete continuous path as the farmland planning analysis result.

[0039] S4.1. Based on bidirectional optional paths, generate independent nodes and mark them as optional or non-optional to form an edge set. Based on the node and edge set, form a directed weighted graph. Specifically, each sub-region is treated as an independent node for both directional paths, and a node set is constructed. For any two nodes, query their connection path structure marker; if it is optional, add the connection path structure to the edge set and record the total length of the connection path as the connection cost; if it is marked as not optional, ignore the connection directly. Construct a directed weighted graph based on the set of nodes and the set of edges; It should be noted that in the actual construction process, the node numbering and edge data structure need to be managed uniformly. For example, a unique identifier can be assigned to each node, and the starting point number, ending point number, and weight of the edge can be recorded in the edge set to form a standard graph data structure. In addition, the consistency between nodes and edges needs to be ensured during the construction process, that is, the starting point and ending point of all edges must exist in the node set.

[0040] S4.2 Select a node as the starting node for access, generate an initial path sequence, perform edge swapping and expansion to obtain the complete path; Specifically, a node is selected as the starting node from the directed weighted graph. Starting from the current starting node, the node with the lowest connection cost to the current starting node and marked as optional is selected from all unvisited nodes as the next node to be visited. The sub-region corresponding to this node is then marked as visited. This process is repeated until all sub-regions have been visited once, resulting in the initial path sequence. Based on the initial path sequence, select two non-adjacent edges in the path and swap them. If both new edges are selectable after the swap and the total path length decreases, then the swap operation is performed; otherwise, the swap operation is not performed. Repeat the swap operation until the total path length no longer decreases, and the final optimized path is obtained. Based on the final optimized path, the path is expanded into the actual execution path, including path segments within sub-regions and connecting path segments between sub-regions, forming a complete path; It should be noted that when selecting a node as the starting node from a directed weighted graph, the forward path state corresponding to the sub-region with the smallest number should be preferred. Furthermore, since each sub-region corresponds to two state nodes (forward and reverse), after visiting one node, the other state node of that sub-region should be excluded from subsequent selections to ensure that each sub-region is visited only once. The "edge" in the context of exchanging two non-adjacent edges in a path refers to the connection between two adjacent nodes in the path. For example, in a path sequence, nodes A→B and C→D each constitute two edges. Secondly, the exchange is to gradually optimize the path, continuously reducing its length. For example, replacing the original connections A→B and C→D with A→C and B→D creates a new path connection structure. Furthermore, the total path length is calculated by summing the connection costs between all adjacent nodes in the path sequence and adding the length of the covered paths within each sub-region, thus obtaining the total length of the complete path.

[0041] S4.3. Based on the complete path, perform partitioning, calculate the actual path length and straight-line distance, and then perform a ratio calculation to obtain the path quality value; Specifically, the complete path is divided into multiple continuous path segments, where each path segment corresponds to a sub-region or a connecting path; For each path segment, calculate the actual path length and the straight-line distance between its starting and ending points; The ratio of the actual path length to the straight-line distance is taken as the path quality value; Set a quality threshold. If the path quality value is greater than the quality threshold, the path segment is determined to be a low-quality path segment; otherwise, the path segment is determined to be a high-quality path segment. Based on low-quality path segments, the access order is swapped within the set of sub-regions involved in the path segment, and the path direction is switched (forward and reverse directions are interchanged). The total length of the connecting path is recalculated, and the scheme with the shortest total length of the connecting path is selected for replacement; this process continues until all path segments meet the quality threshold. By sequentially splicing sub-region coverage paths, connecting paths, and sub-region coverage paths, all sub-region paths are connected into a complete continuous path, and then the complete continuous path is used as the result of farmland planning analysis. It should be noted that: to calculate the actual path length, the path point sequence of the path segment should be obtained first. The path point sequence consists of multiple two-dimensional coordinate points arranged in sequence. For any two adjacent points in the sequence, the distance between the two points is obtained by calculating the square root of the sum of the squares of the differences between the x-coordinates and y-coordinates of the two points. The distances between all adjacent points are then summed to obtain the actual path length of the path segment. As for the straight-line distance, the coordinates of the starting point and the ending point of the path segment can be obtained, and the straight-line distance can be obtained by calculating the Euclidean distance between these two coordinate points. Secondly, the path quality value reflects the "degree of detour" of the path. For example, the closer the path is to a straight line, the closer the ratio is to 1; when the path has many turns or detours, the ratio will increase significantly. This ratio can be used to quantitatively evaluate the quality of a path segment. It should also be noted that the quality threshold is used to limit the degree of detour allowed for a path segment, and its value range should be determined based on the characteristics of agricultural machinery operation, farmland scale, and path planning accuracy requirements. For example, in one implementation, the quality threshold is set to a range of 1.05 to 1.50. If it is close to 1.05, it meets the requirement of a relatively straight path, which is suitable for farmland scenarios with regular plots, few obstacles, and high requirements for operational efficiency. If it is close to 1.50, it allows for a certain degree of path detour, which is suitable for farmland scenarios with complex plot shapes, obstacles, or large restrictions on turning. In actual operation of agricultural machinery, when the path length exceeds 20% of the straight distance, it will significantly increase fuel consumption and operation time. Considering the unavoidable turning paths under curvature constraints and ensuring that the threshold can achieve a balance between path feasibility and path efficiency, 1.20 can be taken as the default value.

[0042] This embodiment also provides a farmland planning and analysis system based on satellite data, including: The farmland closed polygon generation module is used to acquire multi-satellite image data of the target farmland area, perform preprocessing and semantic segmentation, construct a binarized map and fit the endpoints of line segments to form farmland closed polygons. The agricultural machinery parameter acquisition module is used to acquire agricultural machinery parameters and form constraints. The path generation and constraint module is used to perform region decomposition on closed polygons of farmland based on constraints, forming multiple sub-regions with continuous operation attributes, constructing bidirectional selectable coverage paths and state sets, and determining the directional consistency of path connection relationships between sub-regions for coverage paths. The evaluation and reconstruction module is used to construct a directed weighted graph based on the judgment results to determine the access order and coverage direction of the sub-regions, generate a complete path for division and evaluation, and then perform local path reconstruction based on the evaluation results to obtain a complete continuous path as the farmland planning analysis result.

[0043] In summary, by introducing an adaptive scanning direction determination mechanism based on covariance feature analysis, the main extension direction can be automatically extracted according to the shape of the farmland boundary, thereby effectively reducing the number of turns in the path generation process, improving the overall path continuity and operational efficiency. Furthermore, by using a region decomposition method based on the variation of scan line intervals, complex farmland is divided into structurally consistent sub-regions, ensuring good continuity and executability of the paths within each sub-region. In addition, a curvature constraint mechanism is introduced during the sub-region connection process, and the connection path is constructed in conjunction with the minimum turning radius of agricultural machinery, ensuring the executability of the generated path in actual agricultural machinery operations. The construction of a ratio evaluation mechanism between path length and straight-line distance allows for quantitative analysis of path quality, and local path reconstruction is performed based on the evaluation results, thereby effectively reducing path detours and optimizing the overall path length. Therefore, this invention achieves significant improvements in path continuity, executability, and path quality control, and significantly enhances the intelligence level of farmland planning and analysis.

[0044] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A farmland planning and analysis method based on satellite data, characterized in that: include, Multi-satellite imagery data of the target farmland area is acquired, preprocessed, and semantically segmented. A binarized map is constructed and the endpoints of line segments are fitted to form a closed polygon of farmland. Obtain agricultural machinery parameters and formulate constraints; Based on constraints, the closed polygon of farmland is decomposed into multiple sub-regions with continuous operation attributes. A set of bidirectional selectable coverage paths and states is constructed, and the path connection relationship between sub-regions is constructed for the coverage path to determine the directional consistency. Based on the judgment results, a directed weighted graph is constructed to determine the access order and coverage direction of the sub-regions. Then, a complete path is generated for partitioning and evaluation. Based on the evaluation results, a local path reconstruction is performed to obtain a complete continuous path as the farmland planning analysis result.

2. The farmland planning and analysis method based on satellite data as described in claim 1, characterized in that: The preprocessing and semantic segmentation of the multi-satellite imagery data of the target farmland area are performed as follows: Multi-satellite imagery data was acquired through a GIS platform, and coordinate transformation and resampling were performed. The imagery data was then normalized and overlaid to obtain feature vectors. The multi-satellite imagery data included Sentinel-2 multispectral imagery, Landsat-8 imagery, and Gaofen series GF satellite imagery. Based on the feature vector, the probability of farmland category is obtained; farmland category includes farmland, water body, uncultivated land, and shrubland.

3. The farmland planning and analysis method based on satellite data as described in claim 2, characterized in that: The process of constructing a binarized map and fitting the endpoints of line segments to form a closed polygon of farmland is as follows: Based on probability comparison, the binarized values ​​required for the binarized image are generated through a preset judgment threshold. The binarized image is then constructed, and contours are found to form a continuous set of contour points. Based on a continuous set of contour points, the vertical distance is calculated, and then a set of polygon vertices is generated and connected using a preset deviation threshold to form a closed polygon.

4. The farmland planning and analysis method based on satellite data as described in claim 1, characterized in that: The process of obtaining agricultural machinery parameters and forming constraints involves finding the minimum turning radius and operating width of the agricultural machinery in the agricultural machinery design book, using the minimum turning radius as the basic parameter for curvature constraints in path planning, generating constraint formulas, and defining safe operating distances based on the operating width of the agricultural machinery and actual farmland operation requirements.

5. The farmland planning and analysis method based on satellite data as described in claim 1, characterized in that: The specific steps for performing region decomposition on the closed polygon of farmland are as follows: Determine the scanning direction; Scan lines are generated along the scanning direction to perform intersection detection. If they intersect, the intersection coordinates are calculated, sorted, and paired to form several intervals. Otherwise, the intersection coordinates are not calculated, sorted, or matched. Based on several intervals, perform interval segmentation to generate sub-regions.

6. The farmland planning and analysis method based on satellite data as described in claim 1, characterized in that: The construction of the bidirectional selectable coverage path and state set is as follows: Within the sub-region, coverage path segments are generated again according to the scanning direction and connected back and forth to form a bidirectional optional path for continuous path judgment. Extract the positions of the start and end points and their corresponding path directions from the bidirectional optional paths, and construct a state set, including a forward path state set and a reverse path state set.

7. The farmland planning and analysis method based on satellite data as described in claim 1, characterized in that: The determination of directional consistency in the path connection relationships between sub-regions constructed based on the coverage path is as follows: Based on the state set, the path endpoint state of the previous sub-region and the path start state of the next sub-region are obtained as candidate connection state pairs, and connection paths are generated under the constraints of the constraints. After enumerating candidate paths and selecting the optimal connection path as the edge weight, a connection relationship structure is generated, and optional and non-optional markings are performed based on a preset consistency threshold.

8. The farmland planning and analysis method based on satellite data as described in claim 1, characterized in that: After constructing the directed weighted graph to determine the access order and coverage direction of the sub-regions, a complete path is generated for partitioning and evaluation, as follows: Based on bidirectional optional paths, independent nodes are generated and marked as optional or non-optional to form an edge set. Based on the node and edge set, a directed weighted graph is formed. Select a node as the starting node for access, generate an initial path sequence, perform edge swapping and expansion, and obtain the complete path; The path quality value is obtained by dividing the path based on the complete path, calculating the actual path length and the straight-line distance, and then performing a ratio calculation.

9. The farmland planning and analysis method based on satellite data as described in claim 1, characterized in that: The process of performing local path reconstruction based on the evaluation results to obtain a complete and continuous path involves setting a quality threshold. If the path quality value is greater than the quality threshold, the path segment is determined to be a low-quality path segment; otherwise, the path segment is determined to be a high-quality path segment. Based on low-quality path segments, the access order is swapped and the path direction is switched within the set of sub-regions involved in the path segment. The total length of the connecting path is recalculated, and the scheme with the shortest total length of the connecting path is selected for replacement. This process continues until all path segments meet the quality threshold. By sequentially splicing sub-region coverage paths, connecting paths, and sub-region coverage paths, all sub-region paths are connected into a complete continuous path, and then the complete continuous path is used as the result of farmland planning analysis.

10. A farmland planning and analysis system based on satellite data, based on the farmland planning and analysis method based on satellite data according to any one of claims 1 to 9, characterized in that: include, The farmland closed polygon generation module is used to acquire multi-satellite image data of the target farmland area, perform preprocessing and semantic segmentation, construct a binarized map and fit the endpoints of line segments to form farmland closed polygons. The agricultural machinery parameter acquisition module is used to acquire agricultural machinery parameters and form constraints. The path generation and constraint module is used to perform region decomposition on closed polygons of farmland based on constraints, forming multiple sub-regions with continuous operation attributes, constructing bidirectional selectable coverage paths and state sets, and determining the directional consistency of path connection relationships between sub-regions for coverage paths. The evaluation and reconstruction module is used to construct a directed weighted graph based on the judgment results to determine the access order and coverage direction of the sub-regions, generate a complete path for division and evaluation, and then perform local path reconstruction based on the evaluation results to obtain a complete continuous path as the farmland planning analysis result.

Images