Agricultural plot form local updating method and device based on semantic edge and segmentation

Abstract

The invention discloses a semantic edge and segmentation-based agricultural plot form local updating method and device. The method comprises the following steps of: preparing a plot boundary vector of a previous period of a specified application area and a later-period target area remote sensing image which finally needs to be locally updated and extracted; selecting and designing a multi-task learning network model capable of simultaneously generating semantic segmentation and semantic edges; using the semantic segmentation and semantic edge multi-task learning model to train a previous-stage plot boundary vector and a later-stage remote sensing image to obtain a multi-task learning model weight, predicting high-resolution remote sensing image data of a target application area, and obtaining a line result and a surface result of agricultural cultivated land of the application area; and performing small pattern spot removal processing on a surface prediction result, comparing results of two-stage semantic segmentation, and selecting updated land parcel elements according to a threshold value. The updated land parcel elements are classified and processed according to the following conditions: 1, common edges do not exist: if the common edges do not exist, an edge vector result generated by a multi-task learning model through later remote sensing is directly adopted; 2, existence of a common edge: in the two-stage remote sensing images, the area of the agricultural plot may change (increase or decrease), the two-stage images are compared according to an edge fusion algorithm, the same edge in the two-stage images is reserved, and the changed edge is updated; according to the invention, local updating of the remote sensing image plot is finally realized.

Description

Technical Field

[0001] The present invention belongs to the fields of remote sensing image processing and remote sensing image information extraction, and involves semantic segmentation, semantic edge extraction, and a series of post-processing methods. Specifically, the present invention is a method and device for locally updating the morphology of agricultural plots in high-resolution remote sensing images by combining semantic edge extraction and semantic segmentation. Background Art

[0002] Agricultural plots are the material basis for human survival. With the rapid development of remote sensing technology, their spatial information has many important applications, such as crop condition monitoring, risk analysis, and other agricultural development plans, which play an indispensable role in improving agricultural productivity. Due to frequent human intervention, the morphology and boundaries of plots will change locally, especially in scenarios such as land subdivision and reclamation development. This has led to more development needs for local update methods of agricultural plots. For the update extraction of remote sensing ground objects, existing methods mostly rely on change detection to directly extract the changed agricultural plot areas without retaining the boundary information in the original vector. In addition, at present, the edge extraction effect of the model for complex agricultural plots does not exceed manual annotation and is affected by complex remote sensing backgrounds and noises. Therefore, the present invention proposes to update the vector information of the locally updated agricultural plots while retaining the vector information of the agricultural plots that have not been updated.

[0003] Currently, the mainstream algorithms for plot change update are divided into two categories: methods based on end-to-end direct detection and methods based on feature extraction and subsequent analysis. The methods based on feature extraction and subsequent analysis first extract the key features of the data through feature extraction techniques, and then perform change updates through subsequent analysis methods (such as classification, clustering, statistical detection). For example, methods based on change trend analysis, such as trend detection in time series data, or methods for detecting the statistical characteristics of changes. The end-to-end change detection algorithms generally use neural networks to directly detect changes from the input data. Usually, no preprocessing or feature extraction is required. By directly inputting the original data, the change area or result is directly output through the model. For example, HANet uses a lightweight self-attention mechanism and designs a discriminative siamese network to integrate multi-scale features for change detection. With the development of neural networks, the latest network models have also begun to be applied to remote sensing change updates, such as transformers and graph convolutional neural network GCN. The extraction and classification accuracy of agricultural plots has reached a relatively high level, and the ability to understand images has gradually increased. This also gives the end-to-end local update method based on neural networks the significance of further exploration. Summary of the Invention

[0004] The present invention aims to overcome the above-mentioned drawbacks of the prior art and provides a method and device for local updating of high-resolution remote sensing plot based on semantic edges.

[0005] The purpose of the present invention is to solve the problem of local updating of agricultural plots in remote sensing images. Considering the different advantages of semantic segmentation and semantic edge technologies, the present invention specifically uses a deep learning edge detection and segmentation joint task model to learn the features of agricultural plots in remote sensing images, and then predicts the agricultural plot elements in the area where agricultural plot changes need to be detected, outputs an image containing all agricultural plot information in the later period, and performs preliminary post-processing on the segmentation result. Based on the threshold, the updated plots are found on the segmentation result, and the edges of the changed agricultural plots are locally updated.

[0006] A method for local updating of agricultural plot morphology based on semantic edge and segmentation of the present invention is as follows:

[0007] Step 1: Prepare the dataset. Select the image data of the target application area that finally needs change detection, and select samples, which need to contain a large number of dense agricultural plots.

[0008] Step 1.1: Obtain high-resolution remote sensing images. Select high-resolution remote sensing images with higher spatial resolution, which can clearly show the details of some fragmented agricultural plots, and the corresponding image change area does not exceed 10%. Only the samples that meet the requirements can be trained using the vector labels of the previous period.

[0009] Step 1.2: Make remote sensing image samples. Crop the high-resolution remote sensing images into 1000*1000. And crop the vectors correspondingly for training in the following steps.

[0010] Step 1.3: Divide the sample set. Divide the cropped dense remote sensing plot samples, and divide the samples into a training set and a test set according to a certain proportion.

[0011] Step 2: Select a multi-task learning model for semantic segmentation and semantic edge according to the characteristics of agricultural plots in high-resolution remote sensing images, and modify the network to make it more suitable for the plot extraction task in the high-resolution remote sensing images required by the present invention.

[0012] Step 2.1: Selecting a multi-task model instead of a semantic segmentation or semantic edge model is to ensure that the extraction effects of semantic segmentation and semantic edge are consistent and reduce the influence of model performance on the experiment.

[0013] In semantic edge extraction, in order to cope with the problem of unbalanced distribution of edge elements and non-edge pixels in the image, an edge ratio parameter β can be introduced to reduce the influence of this imbalance on network training.

[0014] L=-βlogPr(yi = 1) - (1 - β) ∑ log Pr(y j = 0) #(1)

[0015] Step 2.2: After determining the model, adjust the depth and branches of the network according to the requirements of the agricultural plot extraction task. After determining the network architecture, adjust the hyperparameters to improve the performance of the model in the agricultural plot extraction task.

[0016] Step 3: Use the improved semantic segmentation and semantic edge multi-task extraction network model for agricultural plots designed in Step 2 to train the high-resolution remote sensing image data prepared in Step 1 to obtain a multi-task learning network model, and then use the classified test data in Step 1 to evaluate the model. According to the evaluation results, decide whether to repeat Step 2 to fine-tune the network model structure and parameters to obtain the final optimal extraction model.

[0017] Step 4: Use the trained network model in Step 3 to predict the high-resolution remote sensing image of the target application area prepared in Step 1 to obtain the edge results and segmentation results of the dense agricultural plots in the application area. And perform the operation of removing patches on the segmentation results. Then set a certain threshold for the overlap degree of semantic segmentation, and screen the agricultural plots that need local update according to the threshold.

[0018] Step 4.1: Input the high-resolution remote sensing image prepared in Step 1 into the network model for prediction, and obtain the semantic segmentation result and semantic edge result at the same time.

[0019] Step 4.2: For the prediction results in Step 4.1, including the semantic segmentation result and the edge intensity map generated by the semantic edge. For the segmentation result, set a certain threshold according to the number of points constituting the surface feature, and perform the operation of removing patches according to the threshold to reduce the influence of noise.

[0020] Step 4.3: The present invention uses the Vector Difference method to calculate the geometric difference between two-phase vector maps for the identification of updated areas. This method is simple to calculate and can achieve small errors for relatively complex remote sensing plots such as agricultural plots. The specific method is as follows:

[0021] 1. Calculate the vector difference: By calculating the difference set of the two-phase agricultural plot areas, the updated area is obtained.

[0022] ChangeArea = (Seg 1 - Seg 2 ) ∪ (Seg 2 - Seg 1 )

[0023] Where Seg 1 and Seg 2They are the segmentation map areas of two phases of agricultural plots respectively.

[0024] 2. Set the threshold: The threshold is set by the ratio of the area of the changed region to the total area. If the ratio of the updated area to the total area exceeds a certain threshold (e.g., 10%), it is considered that the region has been updated.

[0025] Step 5: For the agricultural plots that are determined to have been updated in Step 4, for the agricultural plots that have not been updated, directly adopt the previously manually marked plot boundaries; for the agricultural plots that have been updated, adopt the edge results of the later agricultural plots, and use the boundary fusion algorithm to adjust the local boundaries.

[0026] Step 5.1: Distinguish two cases of local updates: 1. No common edge: In the case of no common edge, it is manifested as an increase or decrease in the results of the independent plot segmentation surface. In this case, directly adopt the edge vector results generated by the later remote sensing through the multi-task learning model. 2. There is a common edge: In the two-phase remote sensing images, the area of the agricultural plot may change (increase or decrease). According to the edge fusion algorithm of the present invention, the two-phase images are compared, the same edges in the two-phase images are retained, and the changed edges are updated.

[0027] Step 5.2: Boundary fusion algorithm. Traverse each line segment on the plot vector elements, and store these points in an N x 2 (x, y) tensor. By setting a dynamic threshold according to the surrounding intensity values around the corresponding points in the edge intensity map, and taking the common point with the most consistent intensity value and coordinates, the line segments that have not been updated in the elements are determined. For the updated part, the starting point and the ending point continue to use the points in the updated part of the previous vector, and the updated part adopts the result of vectorizing the line segment in the later edge intensity map. Then, morphological operations such as dilation and erosion are used to further smooth the edges, reduce breaks and discontinuities, improve the continuity of the edges, and finally obtain the edge vector map of the new-phase remote sensing image.

[0028] The second aspect of the present invention relates to a device for local update of the morphological form of agricultural plots based on semantic edges and segmentation, including a memory and one or more processors. Executable code is stored in the memory. When the one or more processors execute the executable code, it is used to implement the method for local update of the morphological form of agricultural plots based on semantic edges and segmentation of the present invention.

[0029] The third aspect of the present invention relates to a computer-readable storage medium, on which a program is stored. When the program is executed by a processor, it implements the method for local update of the morphological form of agricultural plots based on semantic edges and segmentation of the present invention.

[0030] The agricultural plot edge detection method based on semantic segmentation and semantic edge proposed by the present invention. This method can quickly locate the changed plots through semantic segmentation, and finally achieve the precise local update of agricultural plots through semantic edges.

[0031] Compared with the previous update extraction methods, the present invention has the advantage of high precision. In high-resolution remote sensing images, agricultural plots are significantly different from other plots. These plots usually have dense boundaries, individual plots are closely adjacent, the boundary coincidence degree is high, and they are mostly clustered in patches. Since the local update method based on semantic segmentation extracts the segmentation results, it cannot accurately judge the local update of the target plot of agricultural plots on the vectorized post-processed image. Therefore, the present invention finally realizes the local update result in the form of edges. Another advantage of the present invention is the efficiency issue. From the previous change detection, the proportion of changed agricultural plots in the whole image is generally much smaller than the unchanged area. If the two-phase high-resolution remote sensing images are directly extracted, it will waste computing power to a certain extent. Therefore, the present invention first processes the two-phase remote sensing images through a semantic segmentation algorithm, extracts agricultural plot elements, then compares the extracted elements to identify the updated areas, and further performs semantic edge detection on these areas to accurately extract the boundaries of the changed areas.

[0032] Based on the semantic segmentation and semantic edge methods of deep learning, the present invention reuses the existing data and network models to further improve the accuracy and efficiency of the local update of agricultural plots.

[0033] Due to the adoption of the above technical methods, the present invention has the following advantages and beneficial effects:

[0034] 1. By combining the semantic segmentation and semantic edge extraction tasks, the present invention can accurately extract the local updated plots in a complex high-resolution remote sensing environment, solving the problems especially in the cases where there are complex boundaries between agricultural plots and other ground objects, complex boundaries inside agricultural plots, and a high boundary coincidence rate between multiple plots. And it reduces the error caused by the segmentation results. Applying semantic edge detection to the segmentation results can accurately delimit the boundaries of the changed areas.

[0035] 2. The present invention first performs preliminary semantic segmentation and semantic edge extraction on agricultural plots, compares the segmentation results, calculates the differences of the vector maps and sets thresholds to screen the updated areas, and only updates the edges of the updated plots. Compared with the traditional method of performing edge detection on the whole image, it reduces the waste of computing resources and reduces the problems of edge blurring and false extraction caused by model extraction at the present stage, making the best use of the advantage of high accuracy of manual annotation.

[0036] 3. Through morphological operations, the edges of the unupdated and updated regions are fused, ensuring precise boundaries for the updated region and smooth boundaries for the unupdated region. This method improves the handling of boundary details while avoiding problems of boundary discontinuity or inconsistency that may occur when processing regions individually. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 is a flowchart of the method of the present invention;

[0038] Figure 2 is an example of reduction of plot elements when there are no common sides between agricultural plots

[0039] Figure 3 is an example of increase of plot elements when there are no common sides between agricultural plots

[0040] Figure 4 is an example of changes in plot elements when there are common sides between agricultural plots DETAILED DESCRIPTION OF THE INVENTION

[0041] To better understand the specific content of the present invention, the present invention will be introduced in more detail below in conjunction with the drawings and embodiments.

[0042] (1) The main tasks in the data preparation stage include preparing high-resolution remote sensing image data with a resolution of 0.5 m, and selecting data images of the application area for training.

[0043] (1.1) Select the final application area of the method and download the high-resolution remote sensing image data of the application area.

[0044] (1.2) According to the selected high-resolution remote sensing data images, select areas where agricultural plots are relatively dense, and the image change area does not exceed 10%. The later training data uses the image of the later period and the label data of the previous period.

[0045] (1.3) Divide the training set and the test set, and divide the labeled remote sensing images, and divide the samples according to a ratio.

[0046] (2) Select a multi-task learning model according to the characteristics of agricultural plots in the high-resolution remote sensing images. The present invention selects the MFFE model and modifies the network to make it suitable for the plot extraction task in the high-resolution remote sensing images required by the present invention.

[0047] (2.1) Select a multi-task learning model for semantic segmentation and semantic edges, and generate segmentation and edge results simultaneously. In semantic edge extraction, in order to cope with the problem of unbalanced distribution of edge elements and non-edge pixels in the image, an edge ratio parameter β can be introduced to reduce the impact of this imbalance on network training.

[0048] L = -β log Pr(y i = 1) - (1 - β) ∑ log Pr(y j = 0) #(1)

[0049] (2.2) After determining the model, adjust the depth and branches of the network according to the requirements of the agricultural plot extraction task. After determining the network architecture, adjust the hyperparameters to improve the performance of the model in the agricultural plot extraction task.

[0050] (3) Use the improved multi-task learning model designed in step 2 for semantic segmentation and semantic edge extraction applicable to agricultural plots to train the high-resolution remote sensing image training data prepared in step 1, and obtain the network weights of the multi-task learning model for agricultural plots. The number of iterations of the model can be continuously set according to the fitting iteration times during testing, and the batch_size is set according to the size of the dataset, as well as the network performance and computer performance, and should not be too small or too large.

[0051] (4) Use the network model weights trained in step 3 to predict the high-resolution remote sensing image data prepared in step 1, obtain the segmentation surface result and edge line result of the agricultural plot, perform the operation of removing patches on the segmentation result, and screen out the agricultural plots that need to be locally updated according to the threshold.

[0052] (4.1) Transform the large high-resolution remote sensing image in step 1 into a data format that can be input into the multi-task learning model.

[0053] (4.2) Input the large data of the application area into the multi-task learning model to obtain the surface prediction result and line prediction result respectively. The surface result generated by the semantic segmentation branch is grayscale and cannot be directly converted into a vector. Moreover, there are many small patches in the surface result, which will also cause many small fragmented patches in the subsequent vectorization result, affecting the accuracy.

[0054] The formula for the small patch removal algorithm is as follows. block(i) is the i-th block area of the surface result, PIXEL_VALUE represents the grayscale value in the block area, THRESHOLD represents the threshold for the area size of the small patch removal area. When the area is greater than or equal to the threshold, the pixel value of the block area remains unchanged at 255. When the area of the block area is less than the threshold, the pixel value of the block area becomes 0, the same as the background value, and the small patches are removed.

[0055] (4.3) On the vector map, use the Vector Difference method to calculate the geometric difference between two-phase vector maps to identify the updated area. This method is simple to calculate and can achieve small errors for relatively complex remote sensing plots such as agricultural plots. The specific method is as follows:

[0056] 1. Calculate the vector difference: Obtain the updated area by calculating the difference set of the two-phase vector areas of the agricultural plots.

[0057] ChangeArea = (Seg 1 - Seg 2 ) ∪ (Seg 2 - Seg 1 )

[0058] where Seg 1 and Seg 2 are the vector map areas of two phases of agricultural plots respectively.

[0059] 2. Set the threshold: Set the threshold by the ratio of the area of the change area to the total area. If the ratio of the updated area to the total area exceeds a certain threshold (e.g., 10%), it is considered that the area has been updated.

[0060] (5) Determine the agricultural plots that have been updated according to step 4. For the agricultural plots that have not been updated, directly adopt the semantic edges detected and vectorized in the previous period; for the agricultural plots that have been updated, adopt the edge results of the later agricultural plots, and use the boundary fusion algorithm to fuse the edges of the two periods.

[0061] (5.1) Distinguish two cases of local updates: 1. No common edge: In the case of no common edge, it is manifested as an increase or decrease in the result of the independent plot segmentation surface. In this case, directly adopt the edge vector result generated by the multi-task learning model through later remote sensing. 2. There is a common edge: In the two-phase remote sensing images, the area of the agricultural plot may change (increase or decrease). According to the edge fusion algorithm of the present invention, the two-phase images are compared, the same edges in the two-phase images are retained, and the edges that have changed are updated.

[0062] (5.2) Boundary fusion algorithm. Traverse each line segment on the plot vector feature, and store these points in an N x 2 (x, y) tensor. Set a dynamic threshold according to the surrounding intensity values around the corresponding points in the edge intensity map, and take the common point with the most consistent intensity value and coordinates to determine the line segments in the feature that have not been updated. For the updated part, the starting point and the ending point continue to use the points in the updated part of the previous vector, and the updated part adopts the result of vectorizing the line segment in the later edge intensity map. Then, use morphological operations such as dilation and erosion to further smooth the edge, reduce breaks and discontinuities, improve the continuity of the edge, and finally obtain the edge vector map of the new phase of remote sensing image.

[0063] Proved by practice, the local update method of agricultural plot morphology based on semantic edges and semantic segmentation in the present invention can effectively solve the problems of accuracy and efficiency in existing remote sensing ground object extraction methods. By combining the multi-task learning model of deep learning, quickly locate the changed areas of agricultural plots using semantic segmentation, and further accurately extract the boundaries of the changed areas through semantic edges. The present invention can provide more accurate agricultural plot change detection results in high-resolution remote sensing images. In summary, the present invention provides an effective solution for high-precision and high-efficiency remote sensing ground object extraction, and has broad application prospects, especially in the fields of agricultural monitoring, land use change detection, etc.

Claims

1. A local updating method of agricultural land morphology based on semantic edge and segmentation, comprising the following steps: Step 1: Prepare the dataset; select the target application area image data that needs change detection, and select the area containing a large amount of intensive agricultural land; Step 2: Select a multi-task learning model based on the characteristics of agricultural cultivated land in high-resolution remote sensing images, select MFFE, and modify the network to make it suitable for the task of extracting cultivated land from high-resolution remote sensing images; Step 3: Use the improved semantic segmentation and semantic edge multi-task extraction network model suitable for agricultural plots designed in step 2 to train the high-resolution remote sensing image data prepared in step 1 to obtain a multi-task learning network model, and then use the test data classified in step 1 to evaluate the model. According to the evaluation results, decide whether to repeat step 2 to fine-tune the network model structure and parameters to obtain the final optimal extraction model; Step 4: Use the network model trained in step 3 to predict the high-resolution remote sensing image of the target application area prepared in step 1, and obtain the edge results and segmentation results of the dense agricultural plots in the application area; and perform a despeckle operation on the segmentation results; then set a certain threshold for the degree of semantic segmentation overlap, and select the agricultural plots that need to be locally updated according to the threshold; Step 5: According to the agricultural plots that have been updated in step 4, for the agricultural plots that have not been updated, the plot boundaries manually marked in the early stage are directly used; for the agricultural plots that have been updated, the edge results of the later agricultural plots are used, and the boundary fusion algorithm is used to adjust the local boundaries.

2. The method for local updating of agricultural land morphology based on semantic edge and segmentation according to claim 1, characterized in that: Step 1 includes: Step 1.1: Obtain high-resolution remote sensing images; select high-resolution remote sensing images with high spatial resolution, which can clearly show the details of some fragmented agricultural plots, and the corresponding image change area does not exceed 10%. Only samples that meet the requirements can use the vector labels of the previous period for training; Step 1.2: Prepare remote sensing image samples; crop the high-resolution remote sensing image into 1000*1000; and crop the vectors accordingly for training in the following steps; Step 1.3: Divide the sample set; divide the cropped dense remote sensing plot samples into training sets and test sets in proportion.

3. The local updating method of agricultural land morphology based on semantic edge and segmentation according to claim 1, characterized in that: Step 2 includes: Step 2.1: The multi-task model is selected instead of the semantic segmentation or semantic edge model to ensure the consistency of semantic segmentation and semantic edge extraction effects and reduce the impact of model performance on the experiment; In semantic edge extraction, in order to deal with the imbalanced distribution of edge elements and non-edge pixels in images, the edge ratio parameter β can be introduced to reduce the impact of this imbalance on network training; L=-βlogPr(y i =1)-(1-β)∑logPr(y j =0)#(1) Step 2.2: After the model is determined, adjust the depth and branches of the network according to the requirements of the agricultural land extraction task; after the network architecture is determined, adjust the hyperparameters to improve the performance of the model in the agricultural land extraction task.

4. The method for local updating of agricultural land morphology based on semantic edge and segmentation according to claim 1, characterized in that: Step 3 includes: The improved semantic segmentation and semantic edge multi-task extraction network model suitable for agricultural plots designed in step 2 is used to train the high-resolution remote sensing image data prepared in step 1 to obtain a multi-task learning network model. The model is then evaluated using the test data classified in step 1. Based on the evaluation results, it is decided whether to repeat step 2 to fine-tune the network model structure and parameters to obtain the final optimal extraction model.

5. The method for local updating of agricultural land morphology based on semantic edge and segmentation according to claim 1, characterized in that: Step 4 includes: Step 4.1: Input the high-resolution remote sensing image prepared in step 1 into the network model for prediction, and obtain the semantic segmentation result and semantic edge result; Step 4.2: The predicted results in step 4.1 include the semantic segmentation results and the edge strength map generated by the semantic edge; for the segmentation results, a certain threshold is set according to the number of points that make up the surface elements, and despeckle processing is performed according to the threshold to reduce the impact of noise; Step 4.3: Use the VectorDifference method to calculate the geometric difference between the two phases of vector diagrams to identify the update area; specifically, it includes:

1. Calculate vector differences: Calculate the difference between the two agricultural plot areas to obtain the updated area; ChangeArea=(Seg1-Seg2)∪(Seg2-Seg1) Among them, Seg1 and Seg2 are the segmentation map areas of the two phases of agricultural plots; 2. Set the threshold: Set the threshold by the ratio of the area of ​​the changed area to the total area; if the ratio of the updated area to the total area exceeds a certain threshold, the area is considered to have been updated.

6. The method for local updating of agricultural land morphology based on semantic edge and segmentation according to claim 1, characterized in that: Step 5 includes: Step 5.1: Distinguish two cases of local update:

1. No common edge: The case of no common edge is manifested as an increase or decrease in the results of the segmentation of independent plots. In this case, the edge vector results generated by the multi-task learning model of the later remote sensing are directly used; 2. There is a common edge: In the two phases of remote sensing images, the area of ​​the agricultural plot may change. According to the edge fusion algorithm, the two phases of images are compared, the same edges in the two phases of images are retained, and the changed edges are updated; Step 5.2: Boundary fusion algorithm; traverse each line segment on the plot vector element and store these points in an N x 2 (x, y) tensor; set a dynamic threshold based on the surrounding intensity value around the corresponding point in the edge intensity map, take the common point with the most consistent intensity value and coordinate, and then determine the line segment in the element that has not been updated; for the updated part, the starting point and end point continue to use the points in the previous vector updated part, and the updated part uses the result of line segment vectorization in the later edge intensity map; then use morphological operations such as dilation and erosion to further smooth the edge, reduce the phenomenon of breakage and discontinuity, and improve the continuity of the edge, and finally obtain the edge vector map of the new remote sensing image.

7. A local updating device for agricultural plot morphology based on semantic edge and segmentation, characterized in that: It comprises a memory and one or more processors, wherein the memory stores executable code, and when the one or more processors execute the executable code, they are used to implement the local updating method of agricultural land morphology based on semantic edges and segmentation as described in any one of claims 1-6.

8. A computer-readable storage medium, characterized in that: A program is stored thereon, and when the program is executed by a processor, the local updating method of agricultural land morphology based on semantic edges and segmentation as described in any one of claims 1-6 is implemented.

Images