Parcel boundary extraction method, device, equipment and medium

Abstract

The invention discloses a land parcel boundary extraction method, device and equipment and a medium, and relates to the field of remote sensing image processing, and the method comprises the steps: carrying out the land parcel boundary extraction of a satellite remote sensing image of a research region based on a cultivated land parcel boundary extraction model, so as to determine a land parcel boundary image of the research region; the plot boundary image comprises a mask, a contour and a distance; the cultivated land boundary extraction model comprises a residual neural network, a feature enhancement module and a convolution output layer which are connected in sequence; the feature enhancement module comprises a multi-scale feature integration block and a space group attention enhancement mechanism; the residual neural network performs feature extraction of different levels on the input image; the multi-scale feature integration block is used for fusing features of different levels; the space group attention enhancement mechanism carries out space group enhancement on the fused features; and the convolution output layer performs mask prediction, contour extraction and distance estimation on the features after space group enhancement. According to the invention, the plot boundary extraction precision and efficiency are improved.

Description

Technical Field

[0001] The present application relates to the field of remote sensing image processing, and in particular to a method, device, equipment and medium for extracting land parcel boundaries by fusing multi-source high-resolution remote sensing satellite image information. Background Art

[0002] With the continuous growth of the global population and the acceleration of urbanization, new challenges have been raised for food security and land resource management. Accurately extracting the boundaries of cultivated land parcels has become the key to scientific and technological support for land resource planning and agricultural development. Traditional ground observation methods are inefficient and costly in extracting and monitoring the boundaries of cultivated land parcels. Summary of the invention

[0003] The purpose of this application is to provide a method, device, equipment and medium for extracting land parcel boundaries, which can improve the extraction accuracy and efficiency of land parcel boundaries.

[0004] To achieve the above objectives, this application provides the following solutions:

[0005] In a first aspect, the present application provides a method for extracting a land parcel boundary, comprising:

[0006] Obtain satellite remote sensing images of the study area;

[0007] Extracting the plot boundary of the satellite remote sensing image based on the cultivated land plot boundary extraction model to determine the plot boundary image of the study area; the plot boundary image includes a mask, a contour and a distance;

[0008] The farmland plot boundary extraction model is obtained by pre-training with a training data set, wherein the training data set includes a plurality of sample remote sensing images and masks, contour maps and distance maps corresponding to each sample remote sensing image; the farmland plot boundary extraction model includes a residual neural network, a feature enhancement module and a convolution output layer connected in sequence; the feature enhancement module includes a multi-scale feature integration block and a spatial group enhancement attention mechanism;

[0009] Among them, the residual neural network is used to extract features at different levels of the input image; the multi-scale feature integration block is used to fuse features at different levels; the spatial group enhancement attention mechanism is used to perform spatial group enhancement on the fused features; the convolution output layer is used to perform mask prediction, contour extraction and distance estimation on the features after spatial group enhancement.

[0010] In a second aspect, the present application provides a land parcel boundary extraction device, comprising:

[0011] Image acquisition module, used to obtain satellite remote sensing images of the study area;

[0012] A boundary extraction module is used to extract the boundary of the plot from the satellite remote sensing image based on the cultivated land plot boundary extraction model to determine the plot boundary image of the study area; the plot boundary image includes a mask, a contour and a distance;

[0013] The farmland plot boundary extraction model is obtained by pre-training with a training data set, wherein the training data set includes a plurality of sample remote sensing images and masks, contour maps and distance maps corresponding to each sample remote sensing image; the farmland plot boundary extraction model includes a residual neural network, a feature enhancement module and a convolution output layer connected in sequence; the feature enhancement module includes a multi-scale feature integration block and a spatial group enhancement attention mechanism;

[0014] Among them, the residual neural network is used to extract features at different levels of the input image; the multi-scale feature integration block is used to fuse features at different levels; the spatial group enhancement attention mechanism is used to perform spatial group enhancement on the fused features; the convolution output layer is used to perform mask prediction, contour extraction and distance estimation on the features after spatial group enhancement.

[0015] In a third aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-mentioned land parcel boundary extraction method.

[0016] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which implements the above-mentioned land boundary extraction method when executed by a processor.

[0017] According to the specific embodiments provided in this application, this application has the following technical effects:

[0018] The present application provides a method, device, equipment and medium for extracting plot boundaries. The constructed farmland plot boundary extraction model integrates multi-task learning, residual neural network and feature enhancement modules, and especially introduces multi-scale feature integration blocks and spatial group enhanced attention mechanisms, which significantly improve the extraction accuracy and efficiency of plot boundaries, especially when processing complex boundaries and features of different scales, and outperforms other deep learning methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0020] Figure 1 This is an application environment diagram of a land parcel boundary extraction method in one embodiment of the present application;

[0021] Figure 2 A schematic diagram of a flow chart of a method for extracting land parcel boundaries provided in one embodiment of the present application;

[0022] Figure 3 A schematic diagram of a process for generating a contour map and a distance map provided in an embodiment of the present application;

[0023] Figure 4 A schematic diagram of a farmland plot boundary extraction model provided in an embodiment of the present application;

[0024] Figure 5 The ablation experiment results of different models are shown in Figure 2.

[0025] Figure 6 Schematic diagram of the results of boundary extraction for different models;

[0026] Figure 7 A schematic diagram of the functional modules of a land parcel boundary extraction device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0027] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0028] The development of remote sensing technology has provided a new solution for efficient, low-cost and high-temporal and spatial resolution monitoring of cultivated land. The related remote sensing imagery plot boundary extraction methods mainly include traditional machine learning methods and deep learning methods.

[0029] Traditional machine learning methods usually analyze images to distinguish cultivated land by formulating some rules and thresholds or classifying image attributes. The operation is simple but sensitive to noise. With the advancement of technology, traditional machine learning methods are gradually being replaced by deep learning methods, which provide higher automation and adaptability.

[0030] The deep learning method uses a deep learning model based on a convolutional neural network (CNN) to automatically extract plot boundaries. It can maintain high accuracy in complex or changing environments. This technology requires a lot of computing resources and a large amount of training data.

[0031] With the continuous improvement of computing power and the strong support of high-resolution remote sensing images, in the field of remote sensing image plot boundary extraction, the extraction of cultivated land plot boundaries based on deep learning methods is the focus of current research and development by scholars. At the same time, the universality and efficiency of multi-source heterogeneous data fusion and data processing are also constantly enhanced.

[0032] For example, among the convolutional neural networks of deep learning methods, U-Net is a popular deep learning architecture, which is unique in its "U"-shaped structure, which includes a contraction path and a symmetrical expansion path. This design helps the network capture certain details while maintaining contextual information, but when dealing with very small or elongated structures, the model may not be able to fully capture enough details, which may be a limitation for the task of extracting cultivated land boundaries from high-resolution remote sensing images.

[0033] In deep learning methods, the semantic segmentation network (SegNet) uses an encoder-decoder architecture and uses the maximum pooling index upsampling feature map in the encoder stage in the decoder, thereby reducing the number of model parameters and improving segmentation accuracy. However, SegNet is sensitive to noise in images, which affects its performance in practical application scenarios of remote sensing images.

[0034] Multi-scale Attention Network (MA-Net) is an improved deep learning architecture designed for medical image segmentation. MA-Net introduces an attention mechanism to enhance the model's ability to focus on key features, thereby improving its segmentation accuracy. Although the introduction of the attention mechanism improves the ability to extract features, it also increases the complexity of the model, limiting its application in resource-constrained environments. In addition, the optimization and training of MA-Net is more complex than that of standard convolutional networks, requiring more detailed training strategies and parameter adjustments to achieve optimal performance.

[0035] Fully Convolutional Networks (FCN) is the first image segmentation model that replaces the traditional fully connected layer with a convolutional layer. It allows the network to process input images of any size and directly output segmentation maps of the corresponding size, making FCN efficient and flexible in processing image segmentation tasks. However, it has certain limitations in upsampling and feature fusion, and does not fully utilize multi-scale information, resulting in insufficient generalization ability in new scenarios with subtle differences in actual use, and thus performance degradation.

[0036] The Block Extraction-based Multi-Task Network (BsiNet) uses a multi-task learning architecture that emphasizes boundary accuracy during image segmentation. The main task of the model focuses on improving segmentation accuracy, while additional tasks focus on refining boundary processing. However, BsiNet is mainly suitable for specific application scenarios that require accurate boundary analysis. In addition, the structure of the model leads to high computational resource requirements, including significant memory and processing time, which makes it difficult to maintain performance in resource-limited environments.

[0037] ResUNet-a combines the feature extraction capability of the residual network with the effective segmentation structure of U-Net. This structure enhances feature transfer through residual connections, making it perform well when processing images with complex backgrounds and details. However, since the model contains multiple layers of residual blocks and a large number of parameters, it has high requirements for computing resources, and the model's versatility and ease of use are poor in real-time or resource-constrained application scenarios.

[0038] In addition, when processing high-resolution multi-source image data, there are often problems such as insufficient data fusion, low boundary extraction accuracy, and low operating efficiency. For example, multi-source data often includes data of different bands and resolutions, such as high-resolution satellite images and LiDAR (Light Laser Detection and Ranging) data, radar images, etc. In these high-resolution images, the boundaries may be very complex, and may show different characteristics in different bands or sensors, which may make the boundary detection method unable to accurately capture these complex boundaries; similarly, high-resolution multi-source image data is very large, and deep learning methods require huge storage space and computing resources when accessing and calculating its data.

[0039] In response to the problems of insufficient image data fusion, low boundary extraction accuracy and low operating efficiency under multi-source high-resolution image data, this application mainly focuses on designing effective feature extraction, fusion and enhancement, and reducing the waste of resources.

[0040] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0041] The land parcel boundary extraction method provided in the embodiment of the present application can be applied to Figure 1In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be set up separately, or it can be integrated on the server 104, or it can be placed on the cloud or other servers. The terminal 102 can send the satellite remote sensing image of the study area to the server 104. After the server 104 receives the satellite remote sensing image of the study area, it extracts the plot boundary of the satellite remote sensing image based on the cultivated land plot boundary extraction model to determine the plot boundary image of the study area. The server 104 can feed back the plot boundary image to the terminal 102. In addition, in some embodiments, the plot boundary extraction method can also be implemented by the server 104 or the terminal 102 alone.

[0042] The terminal 102 may be, but is not limited to, various desktop computers, laptop computers, smart phones, tablet computers, IoT devices, and portable wearable devices. The IoT devices may be smart speakers, smart TVs, smart air conditioners, smart vehicle-mounted devices, etc. The portable wearable devices may be smart watches, smart bracelets, head-mounted devices, etc. The server 104 may be implemented as an independent server or a server cluster consisting of multiple servers, or may be a cloud server.

[0043] In an exemplary embodiment, Figure 2 As shown, a method for extracting land parcel boundaries is provided. The method is executed by a computer device, and can be executed by a computer device such as a terminal or a server alone, or by a terminal and a server together. In the embodiment of the present application, the method is applied to Figure 1 The server 104 in the example is used for explanation, and the steps include the following steps 201 and 202.

[0044] Step 201, obtaining satellite remote sensing images of the study area.

[0045] Step 202: extract the land parcel boundary of the satellite remote sensing image based on the cultivated land parcel boundary extraction model to determine the land parcel boundary image of the study area. The land parcel boundary image includes a mask, a contour and a distance.

[0046] The cultivated land plot boundary extraction model is obtained by pre-training with a training data set, wherein the training data set includes a plurality of sample remote sensing images and a mask, a contour map and a distance map corresponding to each sample remote sensing image.

[0047] In an exemplary embodiment, the training process of the farmland plot boundary extraction model includes the following steps 301 to 308.

[0048] Step 301, acquiring remote sensing images of a preset area within a plurality of preset time periods to obtain a plurality of preliminary remote sensing images.

[0049] In an exemplary embodiment, resource satellite remote sensing detection information of a preset area within a plurality of preset time periods is first obtained. The resource satellite remote sensing detection information includes geographic information, airspace meteorological environment information and remote sensing images. Then, each remote sensing image is subjected to denoising, cloud removal, geometric correction and orthorectification processing in turn to obtain a plurality of pre-processed images. Then, based on the geographic information, airspace meteorological environment information and pre-processed images corresponding to each remote sensing image, a plurality of preliminary remote sensing images are determined.

[0050] First, the collected remote sensing images are subjected to noise reduction processing to remove random noise in the images. Then, cloud removal technology is applied to process the remote sensing images to remove clouds and shadows that affect the clarity of the images. Then, the geographic distortion of the images is corrected to ensure the geographic location accuracy of the images. After that, the images are converted to the ground coordinate system to eliminate the tilt effect of buildings and terrain. Finally, the images of different bands or different time points are fused to form a complete image of the study area.

[0051] Step 302: performing image fusion on a plurality of preliminary remote sensing images to obtain a fused image.

[0052] Step 303, crop the fused image to obtain multiple cropped images. Specifically, crop the fused image to obtain 20 images of 1000×1000 with the most farmland characteristics and 10 images of 1000×1000 that clearly distinguish farmland from urban lines, ensuring that these images evenly cover the study area. Among them, the characteristics of farmland are identified based on the color of the products produced in the region or the undulation of the farmland ridges observed manually.

[0053] Step 304: Draw the cultivated land plot boundary vector data according to the plurality of cropped images, and perform rasterization processing on the cultivated land plot boundary vector data to obtain a rasterized image of each cropped image. The rasterized image is a mask corresponding to the cropped image.

[0054] Specifically, with the help of 30 cropped 1000×1000 images, high-standard farmland plot boundary vector data were manually drawn using professional GIS software (such as ArcMap) to form high-precision vector data. The drawn vector data was then rasterized through the self-built Python language GDAL script library. The raster data maintained the same resolution and alignment as the original image.

[0055] Step 305 , using overlapping sliding windows to sample the multiple cropped images and the rasterized image corresponding to each cropped image, to obtain multiple sample remote sensing images and a mask of each sample remote sensing image.

[0056] Specifically, the 30 cropped images and their corresponding 30 rasterized images were evenly cropped using overlapping sliding windows, and the entire image was divided into multiple 256×256 remote sensing images to obtain the most original sample set.

[0057] Step 306: Based on the multiple sample remote sensing images, the Euclidean distance transformation formula and the quasi-Euclidean distance transformation formula are used to determine the contour map and distance map of each sample remote sensing image. The contour map and distance map together with the original sample set constitute the model data set.

[0058] Specifically, the Scipy library function is used and the Python code is used to implement the Euclidean distance transformation and the quasi-Euclidean distance transformation. The Euclidean distance transformation formula is:

[0059] The quasi-Euclidean distance transformation formula is:

[0060] Among them, d1 is the Euclidean distance, d2 is the quasi-Euclidean distance, and x i and i are the eigenvectors of two individuals, V is the variance of the corresponding eigenvectors of the two individuals (it should be noted that the variance calculation method here is not divided by n, but divided by n-1), and n is the number of individuals.

[0061] The generation process of contour map and distance map is as follows Figure 3 shown.

[0062] Step 307 , geometrically transform the multiple sample remote sensing images, the contour map and the distance map of each sample remote sensing image to obtain a training data set.

[0063] Specifically, the sample remote sensing images, contour maps, and distance maps are horizontally flipped, rotated 180 degrees and 90 degrees clockwise, and data enhanced without changing the size of the images. Together with the original sample set, they form a training data set and are uniformly and standardizedly named according to information such as the area (rows and columns) from which the images are taken.

[0064] Step 308: Use the training data set to train the farmland plot boundary extraction model to obtain a trained farmland plot boundary extraction model.

[0065] Specifically, the five-fold cross-validation method was used to verify and test the cultivated land parcel boundary extraction model. The cultivated land parcel boundary extraction model with the best round in the test results was used to complete the parcel-level boundary extraction of the study area, and the raster was vectorized and merged to form vector data.

[0066] The effect is evaluated according to the evaluation criteria defined in this application, including Overall Accuracy (OA), Precision (P), Recall (R), Intersection over Union (IoU) and F1 score (F1-score). According to the evaluation indicators, the best performing model parameter configuration is selected to obtain the final boundary prediction results of all sub-images.

[0067] This application continuously learns and improves the ability to extract plot boundaries in multiple iterations of training data sets. Compared with related farmland boundary extraction models based on deep learning methods, this application retains the applicability of extracting plot boundaries from big data remote sensing images, reduces the parameters and amount of calculation of the model, improves the overall performance and computational efficiency of plot boundary extraction, and provides more possibilities and opportunities for the association between spatial geographic information and farmland plot information.

[0068] like Figure 4 As shown in the figure, the farmland plot boundary extraction model includes a residual neural network (ResidualNeural Network, ResNet, a feature enhancement module and a convolution output layer connected in sequence. The feature enhancement module includes a multi-scale feature integration block (Multi-Scale Feature Integration Block, MSFI) and a spatial group-wise enhanced attention mechanism (Spatial Group-wise Enhance, SGE).

[0069] (1) The residual neural network is used to extract features at different levels from the input image.

[0070] In an exemplary embodiment, the residual neural network includes multiple layers of residual blocks, each of which includes multiple convolutional layers and skip connections. Each convolutional layer gradually goes deeper into the image by extracting features at different levels. These feature maps capture the details of the image from the primary to the advanced level, and are output to the subsequent feature enhancement module for processing. The residual neural network uses the following formula to extract features at different levels of the input image:

[0071] x l =x l-1 +F(x l-1 ,W l );

[0072] Among them, x l is the output feature of the l-th layer residual block, x l-1 is the output feature of the l-1th layer residual block, W l is the weight of the residual block of the lth layer, F(xl-1 ,W l ) is the overlap of multiple residual layers, where 0 < l ≤ L and L is the number of layers in the residual block. When l = 0, the output feature of the 0th residual block is the image input to the residual neural network.

[0073] (2) The multi-scale feature integration block is used to fuse features at different levels, which helps the model capture the detailed information of the cultivated land plots.

[0074] The fused feature map M of the (l - 1)th layer l-1 is: M l-1 = MSFI(x l-1 , M l ). First, the deepest feature map extracted by the residual neural network is upsampled (the upsampling factor is 2). The fused feature map of the previous layer is obtained by processing the combination of the layer immediately above the deepest layer of the corresponding residual neural network and M l .

[0075] In an exemplary embodiment, the process of the multi-scale feature integration block fusing features at different levels includes:

[0076] 1) For the lth layer, the deep feature map of the (l + 1)th layer is upsampled to obtain the fused feature map of the (l + 1)th layer. When l = L, the fused feature map of the (L + 1)th layer is the output feature of the Lth residual block.

[0077] 2) The fused feature map of the (l + 1)th layer is combined with the output feature of the lth residual block to obtain the combined feature map of the lth layer.

[0078] Specifically, first, a deep feature map In ∈ R H×W×C and a low-level feature map Res-Layer are input. In is upsampled to obtain the fused feature map In” ∈ R 2H×2W×C . Then, using the concat function, In” ∈ R 2H×2W×C and the low-level feature map Res-Layer are combined into ln' ∈ R 2H×2W×2C .

[0079] 3) The combined feature map of the lth layer is multi-scale fused to obtain the multi-scale fused feature map of the lth layer.

[0080] Specifically, In is multi-scale fused to respectively generate and The formulas are as follows:

[0081]

[0082] where is the fusion feature map of three scales, In' is the feature map after merging the lth layer, In'1 is Conv 1×1 (In') The result after processing, Conv 1×1 Represents a 1×1 convolution operation, Conv 3×3 represents a 3×3 convolution operation, BN represents a batch normalization operation, and (H×W×C') represents the size of the feature map and the number of output channels.

[0083] 4) The fused feature maps of multiple scales in the lth layer are summed element by element and activated to obtain the deep feature map of the lth layer. Specifically, the RELU activation function is used for activation. The multi-scale feature integration is repeated to finally output a 256×256×128 feature map.

[0084] (3) The spatial group enhancement attention mechanism is used to perform spatial group enhancement on the fused features. The spatial group enhancement attention mechanism focuses more on key areas in the image, such as subtle boundary areas, while suppressing irrelevant background information. The formula of the spatial group enhancement attention mechanism is: fused =SGE(M1,x1); where F fused It is the enhanced feature of the spatial group.

[0085] The process of spatial group enhancement by the attention mechanism for spatial group enhancement includes:

[0086] ① Divide the convolution feature matrix Y (i.e., the fused feature) with a length of H, a width of W, and a channel dimension of C into k groups, where Y = {y1, y2, y3, …, y M}, M = H × W, and Represents the feature vector of a specific point in H×W space. The importance of each point is re-determined by calculating the similarity between the feature vector of each point and the global feature vector g, which is obtained by spatially averaging all points:

[0087]

[0088] Where f represents the spatial average function.

[0089] ②Determine y m The importance coefficient c m In order to avoid the influence of coefficient size deviation between different samples, the importance coefficient c m Further normalization, the normalized importance coefficient is obtained

[0090] c m = g·y m =||g||·cos||y m ||·cos(θm );

[0091]

[0092] Among them, θ m Yes m The angle between u and g in the feature space, d is the importance coefficient c m The mean of is the importance coefficient c m The variance of ,∈ is a constant to ensure that the denominator is positive.

[0093] ③Use linear function to adjust To ensure that the normalization can represent the identity transformation and the normalization operation can be restored, the adjusted coefficients are obtained Among them, α and β are function parameters. Here α=β=k=32.

[0094] ④ By adjusting the coefficient p m Apply the Sigmoid function δ, and the derived importance coefficient is the original feature vector y m Scaling is performed to obtain enhanced feature vectors

[0095] ⑤The enhanced feature vector obtained at this time constitutes a 256×256×3 feature map. In the tasks of mask prediction and contour extraction, a softmax classifier is used to obtain the prediction result map, and in the distance map estimation task, a sigmoid classifier is used to obtain the prediction result map.

[0096] (4) The convolutional output layer is used to perform mask prediction, contour extraction, and distance estimation on the features after spatial group enhancement.

[0097] In order to solve the problems of insufficient image data fusion, low boundary extraction accuracy and low computational efficiency under multi-source high-resolution image data, this application adopts a deep learning method strategy and designs a cross-fusion multi-task learning network (CFMT-Net) including three core modules: multi-task learning, feature extraction and feature enhancement. Through this network, a cultivated land parcel boundary extraction model is constructed to improve the data fusion effect and boundary extraction effect of the cultivated land parcel boundary extraction model. The cultivated land parcel boundary extraction model delineates cultivated land from high-resolution satellite images, adopts a multi-task learning method, combined with a residual neural network) and an improved feature enhancement module.

[0098] Among them, the multi-task learning strategy takes mask prediction as the core task, while contour extraction and distance map estimation are used as auxiliary tasks to improve the shape and boundary of the primary task to further refine the prediction of farmland masks and enable the model to obtain better generalization ability and stability.

[0099] Feature extraction through residual neural network not only increases the depth and complexity of feature extraction, including spatial position information, form information and other information in the image, but also reduces the parameters and amount of calculation of the model, making the model have faster calculation speed.

[0100] Afterwards, features at different levels are fused through the feature enhancement module to capture more complex and abstract image features. In this process, the spatial resolution of the image is gradually restored through a series of upsampling layers and jump connections, and features at different levels are combined to restore the image detail information.

[0101] MSFI and SGE are designed and introduced in the feature enhancement module. MSFI can capture and integrate feature information of different regions and scales, reducing the impact of background noise. SGE combines the decoders of three parallel tasks into a single decoder, improving the ability to detect complex boundaries, making the model have fewer parameters and good accuracy, thereby improving the model's ability to detect complex boundaries.

[0102] Finally, the final output image is generated through the convolutional output layer, which makes CFMT-Net produce better generalization and stability than traditional CNNs. During the network training process, the contour extraction and distance map estimation tasks are used to assist in optimizing the main task of mask prediction. In addition, mask and boundary prediction are classification problems, and distance map estimation is a regression problem. During the model training process, the loss function (such as negative log-likelihood loss and regression error loss) is calculated based on the output of the model and the true label, and the Adam optimizer is used to update the network weights to optimize the model performance. This application continuously learns and improves the ability to extract plot boundaries for the training data set in multiple iterative training.

[0103] This application integrates multi-task learning methods during the training process, generates more sample data information from training samples through distance formulas and adds them to model training, and combines residual neural networks and feature enhancement modules to improve the model's generalization ability and plot boundary extraction effects. It improves the accuracy and efficiency of plot boundary extraction to a certain extent, and also provides more possibilities and opportunities for the association between spatial geographic information and cultivated land plot information.

[0104] In another exemplary embodiment, the land parcel boundary extraction method further includes the following steps 203 to 205 .

[0105] Step 203, vectorizing the land boundary image of the study area to obtain vector data of the study area.

[0106] Step 204 , performing superposition analysis on the vector data of the study area and the non-cultivated land vector data, removing non-cultivated land features from the vector data of the study area, and obtaining cultivated land vector data.

[0107] Among them, non-arable land vector data includes water system, road network and urbanized building points. From the perspective of human visual intuition, the first thing observed by satellite is the water system and road network. In addition, in the process of urbanization, there is no large-scale agricultural planting in towns. Based on this, the arcmap image tool is used to overlay the high-resolution image layer, outline the water system (main and branch rivers), road network (main roads, branch routes) and towns (urbanized villages or cities), and erase the overlapping parts. The specific overlay and erasure steps can be implemented in the arcmap tool in sequence.

[0108] Step 205, repairing the boundary breaks in the cultivated land vector data to obtain the cultivated land boundary of the study area. Specifically, repairing the boundary breaks or discontinuities that may appear after the erasure process, and finally obtaining the repaired and accurate cultivated land boundaries of the study area.

[0109] In summary, the beneficial effects of this application include at least the following:

[0110] (1) Improved computational efficiency:

[0111] The data preprocessing process of this application includes: image fusion of multiple resource satellite remote sensing images; cropping images that conform to the study area; drawing the cultivated land boundary vector data of the image; standardizing the sample data, rasterizing the vector data, and enhancing it. Among them, in the case of deep learning training, there is a lack of data, and it is necessary to enhance the sample set data using geometric transformations, including horizontal flipping, 180-degree rotation, and 90-degree clockwise rotation. The size of the image is not changed, and it constitutes an experimental data set together with the original sample set. This application uses rasterized label data, and rasterized data facilitates more spatial analysis and image processing operations, captures pixel-level spatial relationships, ensures data consistency, and improves computational efficiency.

[0112] (2) Reduce the impact of overfitting. Overfitting often occurs in supervised learning methods on model training data, resulting in an inability to generalize to new data. This application uses a multi-task learning method to generate more sample data information from the training data through a distance formula and add it to the model training to improve the generalization ability of the model and make it perform better on new data.

[0113] (3) Consider identifying deeper features and characteristics within the sample. This application uses a residual neural network to increase the network depth, extract more abstract features while reducing the size, and use a feature fusion module to combine features at different levels to capture more complex and abstract image characteristics. These features are crucial for understanding image content and performing accurate plot boundary extraction.

[0114] (5) Design and train a multi-task learning CNN network (mask prediction, contour extraction, and distance map estimation) for collaborative training, requiring the network to be complementary in feature extraction and information processing, and to maintain prediction consistency. Introduce a combination of multiple losses to ensure that the network learns from each other's strengths of different tasks during training and maintains consistency in prediction results, so as to achieve deep collaborative learning.

[0115] (6) Improved operational efficiency: High-resolution multi-source image data is very large, and the method requires a huge amount of storage space and computing resources when accessing and computing its data. This application reduces the amount of computation by utilizing residual connections, and synthesizes different decoders for multiple tasks into a single decoder to complete multi-task learning, optimize network structure and parameter operations to reduce model parameters, and improve overall performance and computing efficiency.

[0116] In order to verify the effectiveness of this application, this application conducted an ablation experiment on CFMT-Net. In the ablation experiment, the baseline model gradually introduced different components to compare and observe the model's processing effect on fine areas such as the boundary between cultivated land and non-cultivated land and small or elongated boundaries, such as Figure 5 As shown, Figure 5 From left to right, the first column is the ablation experiment results of the baseline model, the second column is the ablation experiment results of the baseline model + feature extraction module, the third column is the ablation experiment results of the baseline model + feature enhancement module, the fourth column is the ablation experiment results of the baseline model + multi-task learning, the fifth column is the ablation experiment results of the baseline model + multi-task learning + feature enhancement module, the sixth column is the ablation experiment results of the present application, the seventh column is a schematic diagram of the mask, and the eighth column is the input remote sensing image.

[0117] The baseline model performs well in simple and wide farmland segmentation tasks, but performs poorly in processing small structures and complex boundaries, such as Figure 5 As shown in the first column of , there is an obvious misclassification problem between cultivated land and non-cultivated land.

[0118] After the introduction of the feature extraction module, the baseline model can become clearer and more continuous in the identification of the boundary between cultivated land and non-cultivated land. Figure 5 The continuity of the boundary between cultivated land and non-cultivated land in the first row of the second column is ensured.

[0119] The improved feature enhancement module replaces the decoder of the baseline model, so that the model can reduce over-segmentation when processing the transition area between cultivated land and non-cultivated land, making small boundaries and edges more refined and smoother, such as Figure 5 The thin border line between the 2nd row of the 3rd column and the 3rd row is more refined.

[0120] The fusion of multi-task learning strategies based on the baseline model enhances the model's overall understanding of the image, improves the consistency and accuracy of boundaries, and reduces misjudgment. Figure 5 In the fourth column, the network improves the accuracy of extracting simple cultivated land and subtle areas.

[0121] The model that combines multi-task learning and feature enhancement has broken or continuous points at the elongated boundaries, but it is significantly better than the baseline model that combines a single strategy in ensuring accurate extraction of farmland boundaries. Figure 5 There is a break between the elongated boundary and the simple cultivated land boundary at the center of the image in the 2nd row of the 5th column.

[0122] Finally, CFMT-Net combines all optimization strategies and performs well in extracting farmland boundaries in complex scenes. Its output is highly consistent with the true mask, which shows the accuracy and consistency of the model in processing complex scenes to a certain extent. Figure 5 As shown in the 6th column of .

[0123] The evaluation indicators under different model methods are shown in Table 1. When the input image size is 3×256×256, the floating-point operations per second (GFLOPs) value of CFMT-Net is the smallest (19.41). Compared with other models, it is more efficient in computational efficiency and can complete the same task with fewer computing resources, especially when running on devices with limited resources. Although the CFMT-Net proposed in this application has more model parameters than MA-Net, it is still significantly lower than models such as FCN and ResUNet-a, which reduces the storage requirements of the model and improves the loading speed to a certain extent. This low GFLOPs combined with a reasonable amount of parameters makes CFMT-Net suitable for deployment on platforms with limited storage and memory, and can also meet the needs of high-performance computing environments. It is very suitable for real-time or near-real-time application scenarios that require fast response.

[0124] Table 1 Comparison of evaluation indicators under different model methods

[0125]

[0126] The number of floating-point operations represents the number of gigaflops per second, in G, and the number of parameters represents the complexity of the model, in M, where 1M is equal to 10 6 .

[0127] Table 2 shows the performance comparison of CFMT-Net and other seven convolutional neural network models for farmland boundary extraction on the Jilin-1 image dataset, including the results of overall accuracy, precision, recall, F1 score and intersection-over-union.

[0128] In terms of overall accuracy, CFMT-Net has the highest accuracy, reaching 88.14%, which means that the boundary extraction of this application is more excellent in the accuracy performance of all test samples.

[0129] In terms of accuracy, CFMT-Net has an accuracy of 88.63%, which is higher than the 87.63% of ResUNet-a, and it is better in avoiding mislabeling of non-arable land areas.

[0130] In terms of recall rate, CFMT-Net leads other models, reaching 89.86%, and is very effective in identifying real farmland areas, and can capture all relevant areas to the maximum extent.

[0131] In terms of F1-Score, CFMT-Net leads with 88.95% F1-Score, further confirming the superior performance of the model in accurately identifying farmland boundaries.

[0132] In terms of IoU, compared with the IoU results obtained by U-Net, SegNet, FCN, MA-Net, BsiNet and ResUNet-a, the IoU of CFMT-Net increased by 27.04%, 8.49%, 5.91%, 3.85%, 4.94% and 1.29% respectively, which is better than other models overall and achieves better performance.

[0133] Overall, CFMT-Net outperforms the comparison models on these key performance indicators, especially showing excellent accuracy and consistency in identifying and accurately marking farmland boundaries.

[0134] Table 2 Comparison of experimental results under different model methods

[0135] Model OA (%) P(％) R(％) F1-Score(%) IoU(%) U-Net 72.72 79.77 62.42 62.94 53.33 SegNet 82.33 82.39 84.27 82.90 71.88 FCN 83.88 82.74 87.31 84.72 74.46 MA-Net 85.90 87.71 85.91 85.95 76.52 BsiNet 84.45 83.27 88.71 85.41 75.43 ResUNet-a 87.33 87.46 88.98 87.78 79.08 CFMT-Net 88.14 88.63 89.86 88.95 80.37

[0136] like Figure 6 As shown, the ground-truth masks show the target segmentation regions that each model aims to replicate. These masks are binary, with white areas representing target regions that need to be segmented from the background. Figure 6The first row is the input remote sensing image, the second row is the mask, the third row is the result obtained by this application, the fourth row is the result obtained by ResUNet-a, the fifth row is the result obtained by MA-Net, the sixth row is the result obtained by FCN, the seventh row is the result obtained by SegNet, and the eighth row is the result obtained by U-Net. From the visual comparison, it can be observed that the CFMT-Net prediction results show a high degree of consistency in the correspondence with the real mask, and its segmentation boundaries are clearly defined, and the background is rarely misjudged as the target, showing its high precision and accuracy.

[0137] exist Figure 6 In the first, third, fourth, and fifth columns of the figure, CFMT-Net and ResUNet-a show high accuracy in processing small or elongated region boundaries. However, ResUNet-a has over-segmentation (column 4) and under-segmentation (columns 3 and 5) at the connection points and corners of the elongated boundaries, which makes the boundaries appear slightly rough. In contrast, MANet, FCN, and several other models perform relatively ordinary when processing these subtle boundaries and fail to achieve ideal segmentation results.

[0138] exist Figure 6 In the second, sixth, and seventh columns of the data, we can observe the segmentation results of cultivated land and town, cultivated land and forest, and cultivated land and cultivated land. The prediction results of CFMT-Net are close to the real mask and are very consistent with it, with fewer over-segmentation and under-segmentation problems. In contrast, ResUNet-a misclassifies non-cultivated land areas as cultivated land in some cases, showing a certain degree of over-segmentation, which may be due to the excessive generalization ability of the model in texture and color recognition in forest areas. Although MA-Net conforms to the boundary curve of the real mask, it has relatively low overall accuracy. FCN and SegNet provide good segmentation results in most cultivated land areas. However, at complex land boundary junctions, such as the boundary between cultivated land and town or forest, they often lack sufficient perception ability to accurately divide different land classes. In addition, U-Net often exhibits under-segmentation problems when dealing with continuous or complex terrain changes, and may fail to fully capture small areas of land or continuous terrain changes. These results show the effectiveness of each model in processing image boundaries in practical applications. Visually, it is shown that the CFMT-Net prediction results have a high degree of consistency with the true mask, further proving that CFMT-Net has strong feature extraction and boundary recognition capabilities.

[0139] In summary, this application has been extensively experimentally verified on a public competition dataset (Huawei iFlytek Jilin No. 1 image dataset) and a self-constructed experimental dataset. The experimental results show that the CFMT-Net provided by this application has demonstrated excellent boundary extraction capabilities on the dataset. It has not only achieved remarkable results in improving the accuracy and boundary clarity of the model, but also the model's efficient computing capabilities have opened up new possibilities for the deep association of spatial geographic information with cultivated land plot information and its wide application.

[0140] Based on the same inventive concept, the embodiment of the present application also provides a land parcel boundary extraction device for implementing the land parcel boundary extraction method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more land parcel boundary extraction device embodiments provided below can refer to the limitations of the land parcel boundary extraction method above, and will not be repeated here.

[0141] In an exemplary embodiment, Figure 7 As shown, a land parcel boundary extraction device is provided, including: an image acquisition module 701 and a boundary extraction module 702.

[0142] Among them, the image acquisition module 701 is used to obtain satellite remote sensing images of the study area.

[0143] The boundary extraction module 702 is used to extract the boundary of the plots from the satellite remote sensing image based on the cultivated land boundary extraction model to determine the boundary image of the plots in the study area; the boundary image of the plots includes masks, contours and distances.

[0144] The farmland plot boundary extraction model is pre-trained using a training data set, and the training data set includes multiple sample remote sensing images and masks, contour maps and distance maps corresponding to each sample remote sensing image. The farmland plot boundary extraction model includes a residual neural network, a feature enhancement module and a convolution output layer connected in sequence. The feature enhancement module includes a multi-scale feature integration block and a spatial group enhancement attention mechanism.

[0145] The residual neural network is used to extract features at different levels from the input image. The multi-scale feature integration block is used to fuse features at different levels. The spatial group enhancement attention mechanism is used to perform spatial group enhancement on the fused features. The convolution output layer is used to perform mask prediction, contour extraction and distance estimation on the features after spatial group enhancement.

[0146] In another exemplary embodiment, the land parcel boundary extraction device further includes: a vector processing module 703 , a removal module 704 and a fracture repair module 705 .

[0147] The vector processing module 703 is used to perform vector processing on the land boundary image of the study area to obtain vector data of the study area.

[0148] The removal module 704 is used to perform superposition analysis on the vector data of the study area and the non-cultivated land vector data, remove the non-cultivated land features from the vector data of the study area, and obtain the cultivated land vector data.

[0149] The break repair module 705 is used to repair boundary breaks in the cultivated land vector data to obtain the cultivated land boundary of the study area.

[0150] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0151] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0152] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0153] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0154] In this application, all actions to obtain signals, information or data are carried out in compliance with the relevant data protection laws and policies of the country where they are located and with the authorization given by the owner of the corresponding device.

[0155] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0156] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0157] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0158] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for extracting land parcel boundaries, characterized in that: The land parcel boundary extraction method comprises: Obtain satellite remote sensing images of the study area; Extracting the plot boundary of the satellite remote sensing image based on the cultivated land plot boundary extraction model to determine the plot boundary image of the study area; the plot boundary image includes a mask, a contour and a distance; The farmland plot boundary extraction model is obtained by pre-training with a training data set, wherein the training data set includes a plurality of sample remote sensing images and masks, contour maps and distance maps corresponding to each sample remote sensing image; the farmland plot boundary extraction model includes a residual neural network, a feature enhancement module and a convolution output layer connected in sequence; the feature enhancement module includes a multi-scale feature integration block and a spatial group enhancement attention mechanism; Among them, the residual neural network is used to extract features at different levels of the input image; the multi-scale feature integration block is used to fuse features at different levels; the spatial group enhancement attention mechanism is used to perform spatial group enhancement on the fused features; the convolution output layer is used to perform mask prediction, contour extraction and distance estimation on the features after spatial group enhancement.

2. The land parcel boundary extraction method according to claim 1, characterized in that: The training process of the farmland plot boundary extraction model includes: Acquire remote sensing images of a preset area within a plurality of preset time periods to obtain a plurality of preliminary remote sensing images; Perform image fusion on multiple preliminary remote sensing images to obtain a fused image; Crop the fused image to obtain multiple cropped images; Drawing farmland plot boundary vector data according to the plurality of cropped images, and performing rasterization processing on the farmland plot boundary vector data to obtain a rasterized image of each cropped image; the rasterized image is a mask corresponding to the cropped image; Using overlapping sliding windows to sample multiple cropped images and the rasterized images corresponding to each cropped image, multiple sample remote sensing images and the mask of each sample remote sensing image are obtained; According to multiple sample remote sensing images, the contour map and distance map of each sample remote sensing image are determined using the Euclidean distance transformation formula and the quasi-Euclidean distance transformation formula; Performing geometric transformation on multiple sample remote sensing images, contour maps and distance maps of each sample remote sensing image to obtain a training data set; The training data set is used to train the cultivated land parcel boundary extraction model to obtain a trained cultivated land parcel boundary extraction model.

3. The method for extracting land parcel boundaries according to claim 2, characterized in that: Acquire remote sensing images of a preset area within multiple preset time periods to obtain multiple preliminary remote sensing images, including: Acquire resource satellite remote sensing detection information of a preset area within a plurality of preset time periods; the resource satellite remote sensing detection information includes geographic information, airspace meteorological environment information and remote sensing images; Each remote sensing image is subjected to denoising, cloud removal, geometric correction and orthorectification processing in turn to obtain multiple pre-processed images; According to the geographic information, airspace meteorological environment information and pre-processed images corresponding to each remote sensing image, multiple preliminary remote sensing images are determined.

4. The method for extracting land parcel boundaries according to claim 1, characterized in that: The residual neural network uses the following formula to extract features at different levels from the input image: x l =x l-1 +F(x l-1 ,W l ); Among them, x l is the output feature of the l-th layer residual block, x l-1 is the output feature of the l-1th layer residual block, W l is the weight of the residual block of the lth layer, F(x l-1 , W l ) is the overlap of multiple residual layers, 0<l≤L, L is the number of layers of residual blocks, and when l=0, the output feature of the 0th layer residual block is the image input to the residual neural network.

5. The method for extracting land parcel boundaries according to claim 4, characterized in that: The process of fusing features at different levels by the multi-scale feature integration block includes: For the lth layer, the deep feature map of the l+1th layer is upsampled to obtain the fused feature map of the l+1th layer; when l=L, the fused feature map of the L+1th layer is the output feature of the residual block of the Lth layer; The fused feature map of the l+1th layer is merged with the output feature of the residual block of the lth layer to obtain the merged feature map of the lth layer; The merged feature maps of the lth layer are fused at multiple scales to obtain the fused feature maps of multiple scales of the lth layer; The fused feature maps of multiple scales in the lth layer are summed element by element and activated to obtain the deep feature map of the lth layer.

6. The method for extracting land parcel boundaries according to claim 5, characterized in that: The following formula is used to perform multi-scale fusion on the feature maps after merging the lth layer: in, is the fusion feature map of three scales, In′ is the feature map after merging at the lth layer, Conv 1×1 Represents a 1×1 convolution operation, Conv 3×3 represents a 3×3 convolution operation, and BN represents a batch normalization operation.

7. The method for extracting land parcel boundaries according to claim 1, characterized in that: The land parcel boundary extraction method further comprises: Vectorizing the boundary image of the plot of land in the study area to obtain vector data of the study area; Performing superposition analysis on the vector data of the study area and the vector data of non-cultivated land, removing non-cultivated land features from the vector data of the study area, and obtaining cultivated land vector data; The boundary breaks in the cultivated land vector data are repaired to obtain the cultivated land boundary of the study area.

8. A land parcel boundary extraction device, applied to the land parcel boundary extraction method according to any one of claims 1 to 7, characterized in that: The land block boundary extraction device comprises: Image acquisition module, used to obtain satellite remote sensing images of the study area; A boundary extraction module is used to extract the boundary of the plot from the satellite remote sensing image based on the cultivated land plot boundary extraction model to determine the plot boundary image of the study area; the plot boundary image includes a mask, a contour and a distance; The farmland plot boundary extraction model is obtained by pre-training with a training data set, wherein the training data set includes a plurality of sample remote sensing images and masks, contour maps and distance maps corresponding to each sample remote sensing image; the farmland plot boundary extraction model includes a residual neural network, a feature enhancement module and a convolution output layer connected in sequence; the feature enhancement module includes a multi-scale feature integration block and a spatial group enhancement attention mechanism; Among them, the residual neural network is used to extract features at different levels of the input image; the multi-scale feature integration block is used to fuse features at different levels; the spatial group enhancement attention mechanism is used to perform spatial group enhancement on the fused features; the convolution output layer is used to perform mask prediction, contour extraction and distance estimation on the features after spatial group enhancement.

9. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the land boundary extraction method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for extracting land parcel boundaries described in any one of claims 1 to 7 is implemented.

Images