A land use planning verification method, device, storage medium and equipment

Abstract

This application belongs to the field of land use planning verification technology, and discloses a method, apparatus, storage medium, and equipment for land use planning verification. It involves acquiring a planning vector map and a color block planning map of the area to be verified; inputting the color block planning map into a convolutional neural network to extract color block features, obtaining a multi-scale feature map; inputting the multi-scale feature map into a pre-trained semantic segmentation network for color block segmentation, outputting color block regions; constructing a topological structure map of the area to be verified based on the color block regions; inputting the topological structure map and the planning vector map into a graph neural network to analyze topological relationships, and obtaining a state prediction result for each target node by aggregating the features of the target node itself and the features of its neighboring nodes. The state prediction results include compliance, attribute violation, spatial violation, or combined violation; integrating the state prediction results, and generating a verification report based on a rule engine. This can improve the efficiency of land use planning verification tasks and enhance the consistency and stability of verification results.

Description

Technical Field

[0001] This application relates to the field of land use planning verification technology, and in particular to a method, apparatus, storage medium and equipment for verifying land use planning. Background Technology

[0002] In current land use planning verification work, verification departments mainly rely on traditional Geographic Information System (GIS) platforms, such as ArcGIS, combined with manual interpretation and operation. This model requires technicians to import multi-source heterogeneous spatial data, such as spatial planning vector maps, regulatory detailed planning maps, and industrial block data, manually compare and identify color blocks, extract and associate attribute information item by item, and conduct compliance analysis according to relevant rules. The entire process covers multiple stages, including data import, color block identification, attribute matching, and rule judgment, and is highly dependent on the experience and proficiency of the operators.

[0003] However, this model has gradually revealed several limitations in practical applications. On the one hand, due to its heavy reliance on manual processing, it suffers from low operational efficiency when faced with the verification tasks of hundreds of projects and tens of thousands of plots of land each year. Especially during peak periods, a single verification cycle can last for several weeks, making it difficult to meet the management requirements of rapid response and efficient approval. On the other hand, manual interpretation is greatly affected by the resolution of planning maps, differences in color representation, and personal subjective experience, leading to significant fluctuations in the accuracy of color block recognition and attribute matching, making it difficult to guarantee consistency in judgments, and thus directly affecting the accuracy and reliability of the final verification results. Therefore, existing technical means are increasingly unable to meet the needs of large-scale, high-requirement planning and land use verification work in terms of efficiency, accuracy, and stability. Summary of the Invention

[0004] Therefore, embodiments of this application provide a method, apparatus, storage medium, and device for verifying land use planning, which can improve the efficiency of land use planning verification tasks and enhance the consistency and stability of verification results.

[0005] Firstly, this application provides a method for verifying land use planning.

[0006] This application is achieved through the following technical solution: A method for verifying land use planning includes: Obtain the planning vector map and color block planning map of the area to be verified; The color block planning map is input into a convolutional neural network to extract color block features, resulting in a multi-scale feature map. The multi-scale feature map is input into a pre-trained semantic segmentation network for color block segmentation, and the color block regions are output. Based on the color block regions, construct a topological structure diagram of the region to be verified; The topology diagram and planning vector diagram are input into a graph neural network to analyze the topological relationships. By aggregating the characteristics of the target node itself and the characteristics of the neighboring nodes, the state prediction result of each target node is obtained. The state prediction result includes compliance, attribute violation, spatial violation or compound violation. The state prediction results are integrated, and a structured verification report is generated based on the rule engine.

[0007] In a preferred example of this application, the method may be further configured to obtain a planning vector map and a color block planning map of the area to be verified, and may also include: Use spatial correction tools to correct the planning vector map and color block planning map to a unified coordinate system.

[0008] In a preferred example of this application, the network backbone of the convolutional neural network is selected as ResNet50 as the architecture, and an SE attention module is inserted into ResNet50.

[0009] In a preferred embodiment of this application, the method can be further configured to construct a topological structure map of the region to be verified based on the color block region, including: Extract the geometric and attribute features of the color block regions and establish initial topological relationships; Each color block area is set as a target node, and the control area is extracted from the planning vector map. Each control area is set as a reference vector node. Through spatial analysis and attribute matching, the node relationship between the target node and the reference vector node is quantified. Use spatial analysis tools to determine edge types and edge relationships, and define the relationships between edges in the graph based on the spatial relationships between target nodes and between target nodes and reference nodes; The initial topological relationships are corrected based on the node relationships and edge relationships to generate a topological structure graph for input to the graph neural network.

[0010] Secondly, this application provides a land use planning verification device.

[0011] This application is achieved through the following technical solution: A land use planning verification device, used to perform the land use planning verification method described in the first aspect above, comprising: The data collection module is used to acquire the planning vector map and color block planning map of the area to be verified; The color block recognition module is used to input the color block planning map into the convolutional neural network to extract color block features and obtain a multi-scale feature map; The multi-scale feature map is input into a pre-trained semantic segmentation network for color block segmentation, and the color block regions are output. The graph structure construction module is used to construct a topological structure graph of the area to be checked based on the color block area; The verification module is used to input the topology diagram and planning vector diagram into the graph neural network to analyze the topological relationships. By aggregating the characteristics of the target node itself and the characteristics of the neighboring nodes, the state prediction result of each target node is obtained. The state prediction result includes compliance, attribute violation, spatial violation or compound violation. The report generation module is used to integrate the state prediction results and generate a structured verification report based on the rule engine.

[0012] In a preferred embodiment of this application, the data collection module may further include a preprocessing unit, which is used to correct the planning vector map and color block planning map to a unified coordinate system using a spatial correction tool.

[0013] In a preferred example of this application, the network backbone of the convolutional neural network is selected as ResNet50 as the architecture, and an SE attention module is inserted into ResNet50.

[0014] In a preferred example of this application, the graph structure building module can be further configured to specifically be used for: Extract the geometric and attribute features of the color block regions and establish initial topological relationships; Each color block area is set as a target node, and the control area is extracted from the planning vector map. Each control area is set as a reference vector node. Through spatial analysis and attribute matching, the node relationship between the target node and the reference vector node is quantified. Use spatial analysis tools to determine edge types and edge relationships, and define the relationships between edges in the graph based on the spatial relationships between target nodes and between target nodes and reference nodes; The initial topological relationships are corrected based on the node relationships and edge relationships to generate a topological structure graph for input to the graph neural network.

[0015] Thirdly, this application provides a computer-readable storage medium.

[0016] This application is achieved through the following technical solution: A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of any of the above-described land use planning verification methods.

[0017] Fourthly, this application is achieved through the following technical solution: A computer device includes 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 steps of any of the above-described land use planning verification methods.

[0018] In summary, compared with the prior art, the beneficial effects of the technical solution provided by the embodiments of this application include at least the following: This application obtains a planning vector map and a color block planning map of the area to be verified; inputs the color block planning map into a convolutional neural network to extract color block features to obtain a multi-scale feature map; inputs the multi-scale feature map into a pre-trained semantic segmentation network for color block segmentation and outputs color block regions; constructs a topology map of the area to be verified based on the color block regions; inputs the topology map and the planning vector map into a graph neural network to analyze the topological relationships, and obtains the state prediction result of each target node by aggregating the features of the target node itself and the features of the neighboring nodes, wherein the state prediction result includes compliance, attribute violation, spatial violation or compound violation; integrates the state prediction results and generates a structured verification report based on a rule engine. Covering the entire process from color block recognition, spatial analysis, compliance review to report output, it completely replaces traditional manual operations, reducing the completion cycle of large-scale verification tasks from several weeks to extremely short time, significantly improving the processing efficiency of verification tasks; through deep learning models, it effectively identifies color block areas in drawings of different resolutions, transforms the color block areas into topological structure diagrams, analyzes the topological relationships within them, obtains the compliance attributes between color blocks, and automatically generates standardized verification reports that meet the requirements of the specifications, effectively improving the stability of the recognition results of verification tasks. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating a land use planning verification method provided in one embodiment of this application. Detailed Implementation

[0020] This specific embodiment is merely an explanation of this application and is not intended to limit it. After reading this specification, those skilled in the art can make modifications to this embodiment without contributing any inventive step, but such modifications are protected by patent law as long as they fall within the scope of the claims of this application. To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] Furthermore, the term "and / or" in this application is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this application, unless otherwise specified, generally indicates that the preceding and following related objects have an "or" relationship.

[0022] In this application, the terms "first," "second," etc., are used to distinguish identical or similar items with essentially the same function. It should be understood that there is no logical or temporal dependency between "first," "second," and "nth," nor are there any restrictions on quantity or execution order.

[0023] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0024] The embodiments of this application will now be described in further detail with reference to the accompanying drawings.

[0025] See Figure 1 The illustration shown is a land use planning verification method provided in the first exemplary embodiment of this application, the specific steps of which include: S1: Obtain the planning vector map and color block planning map of the area to be verified.

[0026] In practice, the first step is to determine the boundary of the area to be verified and obtain the planning vector map and the color-block planning map with land use color blocks within that boundary. The planning vector map is a digital map stored in vector data format, representing the spatial planning results. It uses geometric shapes such as points, lines, and polygons, along with their associated attribute data, to accurately describe the spatial location, shape, topological relationships, and properties of planning elements. The associated attribute data includes at least indicators such as plot number, land use code, land area, plot ratio, building density, and green space ratio. Typically, the original planning vector map may be in a specific format, such as gdb, shp, or dwg, requiring professional GIS software for viewing and processing. The color-block planning map is a visual rendering of the planning vector map. Based on the land use boundary layer in the planning vector map, different colors are filled to plots with different land use properties. For example, residential land is filled with yellow, commercial land with red, industrial land with brown, and green space with green. The color-block planning map is usually in high-resolution PDF, JPEG, or PNG formats. To facilitate subsequent processing, the GDAL and OpenCV libraries of Python are used to process the data format, converting the planning vector map into SHP format and the color block planning map into TIFF format.

[0027] Color block types are labeled for color block areas of different land use in the color block planning map. For example, residential land R1 is labeled as 1, industrial land M1 is labeled as 2, and commercial land B1 is labeled as 3, forming labeled datasets with different resolutions of 100dpi, 200dpi, 300dpi, 400dpi, 500dpi, and 600dpi, with 200 samples at each resolution. The labeled datasets are used for training the subsequent color block recognition model.

[0028] S2: Input the color block planning map into the convolutional neural network to extract color block features and obtain a multi-scale feature map.

[0029] The format of the multi-scale feature map is Tensor.

[0030] Specifically, the convolutional neural network backbone uses ResNet50 as its base architecture and inserts the SE attention module into ResNet50.

[0031] Specifically, this application employs an improved CNN model and a semantic segmentation network for land use color block recognition. The improved CNN model is used for color block feature extraction. To address the gradient vanishing problem during deep training of traditional CNNs, the network backbone uses ResNet50 as its architecture. Furthermore, Squeeze-and-Excitation (SE) attention modules are inserted at the end of the 3rd and 4th convolutional blocks of the ResNet50 network to enhance the extraction capabilities of color and boundary features of the color blocks.

[0032] Among them, the SE attention module performs a Squeeze operation on the feature map output by the convolution block, that is, performs global average pooling to obtain a 1×1×C channel feature vector, where C is the number of channels in the feature map, highlighting the channel importance of the color of the color block; The excitation operation involves using two fully connected layers with a ReLU activation function in between, and activating the output layer with a Sigmoid function to generate channel weights. These weights are then applied to the original feature map to enhance the key features of the color blocks.

[0033] The loss function used is the cross-entropy loss function, which addresses the imbalance of similar color block samples in the standard color block dataset. For example, if residential land color blocks account for 40%, commercial land for 30%, industrial land for 10%, and the remainder are other types of land, category weights are introduced, as shown in the following formula: , Where N is the number of samples and C is the number of land use categories. Let c be the category weight of the color patch. Let be the true label of the i-th sample in class c, where the true label is either 0 or 1. Predict the probability that the i-th sample belongs to the c-th color patch for the model.

[0034] When the color block planning map is input into the convolutional neural network, the image pixel values ​​are normalized from [0,255] to [0,1] using the Python OpenCV library to remove extreme noise pixels in the image.

[0035] In this application, an improved CNN model is used as the encoding end of U-Net. Its core is to extract multi-scale feature maps of color, boundary, and spatial distribution of color patches. The multi-scale feature maps contain four levels of encoded features: the output of the first encoding layer is (batch_size, H / 4, W / 4, 256), the output of the second encoding layer is (batch_size, H / 8, W / 8, 512), the output of the third encoding layer is (batch_size, H / 16, W / 16, 1024), and the output of the fourth encoding layer is (batch_size, H / 32, W / 32, 2048). The multi-scale feature maps are compatible with the bilinear interpolation upsampling and skip connection fusion logic of the U-Net decoding end and can be directly used as input to the decoding end.

[0036] To address the coexistence of large, continuous areas and small, scattered areas within land use color blocks, this application employs a multi-scale feature fusion mechanism to simultaneously capture both global spatial layout and local detail boundaries. Low-level features primarily characterize local details of the color blocks, such as boundary pixel gradients and color distribution; high-level features carry more abstract semantic information, such as spatial topological relationships between color blocks and land use category characteristics. Building upon this foundation, an SE attention module is embedded after the 3rd and 4th convolutional blocks of the ResNet50. This module enhances color channels with strong classification discriminative power (e.g., increasing the weight of the red channel in identifying commercial land), enabling the model to focus more on key color features. This provides a more discriminative feature representation for the subsequent semantic segmentation network, directly supporting its accurate land use category determination.

[0037] S3: Input the multi-scale feature map into the pre-trained semantic segmentation network to perform color block segmentation and output the color block region.

[0038] In this application, an improved CNN is used as the encoding end of U-Net, which is mainly responsible for feature extraction; the decoding end adopts bilinear interpolation upsampling, combined with the skip chain structure of the encoding segment, to preserve the detailed information of the color block boundary, complete the pixel-level accurate segmentation, and output the color block region of the raster mask.

[0039] It should be noted that the semantic segmentation network uses the Dice loss function to address the imbalance between background and foreground pixel ratios in color block region segmentation. The formula for the loss function is as follows: , Where H and W are the image height and width, Here are the true segmentation labels for pixels (i, j), where a label of 1 represents a color patch and a label of 0 represents the background. Predict segmentation labels for the model.

[0040] U-Net semantic segmentation outputs a pixel-level raster mask in TIFF format, with different data labels representing different land use types. For example, a data label of 0 represents the background, a data label of 1 represents residential land, and a data label of 2 represents industrial land.

[0041] It should be noted that this application uses a pre-labeled dataset to train the semantic segmentation model, and the Adam optimizer is selected. The initial learning rate is 0.001, which decays to 0.5 every 5 epochs. The training epochs are set to 50 epochs. Training is stopped when the validation set loss does not decrease for 3 consecutive epochs, and the optimal model is saved.

[0042] Test samples with different resolutions (100dpi, 200dpi, 300dpi, 400dpi, 500dpi, and 600dpi) were used to validate the recognition metrics of the convolutional neural network and semantic segmentation model. The recognition metrics included accuracy, recall, and F1 score. For input data with a resolution of 100dpi, the accuracy was 92.3%, recall was 91.5%, F1 score was 91.9%, and single-image processing time was 0.8s. For input data with a resolution of 200dpi, the accuracy was 94.7%, recall was 93.9%, F1 score was 94.3%, and single-image processing time was 1.2s. For input data with a resolution of 300dpi, the accuracy was 96.5%, recall was 95.8%, F1 score was 96.1%, and single-image processing time was 1.5s. For input data with a resolution of 400 dpi, the accuracy was 97.2%, the recall was 96.9%, the F1 score was 97.0%, and the processing time per image was 2.1 s. For input data with a resolution of 500 dpi, the accuracy was 97.5%, the recall was 97.1%, the F1 score was 97.3%, and the processing time per image was 2.8 s. For input data with a resolution of 600 dpi, the accuracy was 97.8%, the recall was 97.5%, the F1 score was 97.6%, and the processing time per image was 3.5 s. The results show that the color block recognition method of this application achieves an accuracy greater than 92% at resolutions ranging from 100 dpi to 600 dpi, meeting the resolution adaptability requirements of planning and verification. The processing time per image is less than 3.5 s, effectively improving processing efficiency.

[0043] S4: Based on the color block regions, construct a topological structure diagram of the region to be checked.

[0044] After obtaining the color patch regions, these regions are converted into a topological graph that can be processed by a graph neural network. This mainly involves post-processing of the grid mask, extraction of color patch region information, construction of GNN nodes and edges, and finally outputting the graph structure.

[0045] S5: Input the topology diagram and planning vector diagram into the graph neural network to analyze the topological relationships. By aggregating the characteristics of the target node itself and the characteristics of the neighboring nodes, the state prediction results of each target node are obtained. The state prediction results include compliance, attribute violation, spatial violation or compound violation.

[0046] Specifically, firstly, ensure that the spatial reference of the topology map and the planning vector map are consistent, and that all nodes are unified to latitude and longitude / planar coordinates under the CGCS2000 coordinate system; supplement the control attributes of the planning vector map, such as the prohibited development attribute of the ecological protection red line and the access type of the urban development boundary, into the feature vector of the corresponding reference vector node, while retaining the land use nature and other attributes of the color block nodes to provide attribute support for subsequent relationship judgment; based on ArcGIS spatial analysis tools and geometric calculations, make preliminary judgments on the spatial relationships between color block nodes and reference vector nodes (such as the three zones and three lines), and between color block nodes, to form an initial edge set; Based on purified color block raster data and planning vector data, noise is first eliminated through morphological opening operations, and connected regions are extracted as color block nodes. Simultaneously, key control areas such as the "three zones and three lines" are extracted from the vector map as reference vector nodes. The features of each node are composed of both geometric and attribute information. Subsequently, a heterogeneous graph is constructed with nodes as entities and spatial topological relationships as edges. The neighborhood aggregation capability of a graph neural network (GCN / GAT) is used to optimize the initial relationships—by aggregating neighborhood node features and learning edge weights, it effectively eliminates false connections, identifies contiguous illegal areas, and captures indirect topological associations. Finally, the model outputs a stable and unique set of topological relationships. This result can be directly bound to compliance rules, providing a core basis for automated land use compliance scoring and assessment.

[0047] S6: Integrate status prediction results and generate a structured verification report based on the rule engine.

[0048] Based on a preset template, this application automatically integrates the basic information of the verified land parcels (name, coordinate range, time), the multi-scale accuracy and statistical results of color block recognition, the compliance score including three-dimensional details and violation diagrams, and specific control guidelines (admission conditions / rectification suggestions). Finally, it generates a structured standard report with one click and supports exporting it as a Word document, PDF file or ArcGIS project file to meet the application needs of different scenarios such as approval, archiving and collaborative analysis.

[0049] In some embodiments, obtaining the planning vector map and the color block planning map with land use color blocks for the area to be verified further includes: Use spatial correction tools to correct the planning vector map and color block planning map to a unified coordinate system.

[0050] In practice, the spatial correction tools in ArcGIS are used to correct the planning vector map and color block planning map to the CGCS2000 geodetic coordinate system to ensure accurate alignment in the future.

[0051] Furthermore, the RANSAC algorithm can be used to eliminate geometric distortions in the color block planning map, such as color block boundary offsets caused by scan map tilt.

[0052] In some embodiments, a topological structure map of the region to be verified is constructed based on the color block region, including: Extract the geometric and attribute features of the color block regions and establish initial topological relationships; Each color block area is set as a target node, and the control area is extracted from the planning vector map. Each control area is set as a reference vector node. Through spatial analysis and attribute matching, the node relationship between the target node and the reference vector node is quantified. Use spatial analysis tools to determine edge types and edge relationships, and define the relationships between edges in the graph based on the spatial relationships between target nodes and between target nodes and reference nodes; The initial topological relationships are corrected based on the node relationships and edge relationships to generate a topological structure graph for input to the graph neural network.

[0053] In practice, morphological opening operations are first performed, involving sequential erosion and dilation, to remove isolated noise pixels from the raster mask, such as scan spots and tiny pixel clusters misidentified during segmentation. This smooths the boundaries and preserves continuous color areas. Then, OpenCV's `findContours` function is used to extract the processed connected regions, and a minimum area threshold is set. This minimum area threshold can be adjusted according to the actual verification accuracy requirements to effectively filter out invalid small regions caused by segmentation errors and reduce their impact on GNN analysis. For example, the minimum area threshold can be set to 10 pixels, corresponding to an actual land area of ​​100 square meters.

[0054] For each purified color block region, geometric and attribute features are extracted to jointly constitute the node representation of the graph neural network. Geometric features include center point coordinates, boundary contour coordinate set, color block area, and shape factor, used in the GNN to analyze the spatial topological relationship between the color block and the vector layer, such as determining whether the center point falls within the ecological protection red line and whether the boundaries overlap. Attribute features include land use label, segmentation confidence score, and resolution adaptation marker, mainly used to drive the rule engine to perform land use matching (comparison with control planning data) and compliance scoring (confidence-based correlation identification accuracy score).

[0055] Graph neural network nodes are divided into two types: color block nodes and reference vector nodes. Each purified color block region corresponds to a color block node. The feature vector of a color block node is a concatenation vector of geometric features and attribute features, which can be represented as: [color block node | center point longitude, center point latitude, actual area, shape factor, land use code, segmentation confidence, resolution label]. The feature dimension can be adjusted according to the actual number of features extracted.

[0056] Key control areas such as the three zones and three lines (ecological protection red line, permanent basic farmland, urban development boundary), industrial block line, and ecological corridor are extracted from the planning vector map. Each control area corresponds to a reference node, and the node characteristics are [reference node | center point coordinates, area type, control rule code].

[0057] Spatial analysis tools are used to determine edge types and edge relationships. Based on the spatial relationships between target nodes and between target nodes and reference nodes, the relationships between edges in the graph are defined as: overlapping edges, adjacent edges, and edges far away.

[0058] Specifically, when the ratio of the intersection area of ​​the color block region and the reference vector region to the area of ​​the color block region is greater than 5%, the boundary of the color block region and the boundary of the reference vector region are set as overlapping edges, and the weight is set as the overlap ratio; When the color block region does not intersect with the reference vector region, and the shortest distance between the boundary of the color block region and the boundary of the reference vector region is less than 50 meters, the boundary of the color block region and the boundary of the reference vector region are set as adjacent edges, and the weight is set to the shortest distance. When the color block region does not intersect with the reference vector region, and the shortest distance between the boundary of the color block region and the boundary of the reference vector region is greater than or equal to 50 meters, the boundary of the color block region and the boundary of the reference vector region are set to the far edge, and the weight is set to 0.

[0059] This involves using ArcGIS's Intersect tool to calculate the intersection area and using the Shapely library to calculate the shortest distance between geometric objects.

[0060] After determining the type of an edge, the relationship between the edges is stored using an adjacency matrix. The dimension of the adjacency matrix is ​​the square of the total number of nodes, and the matrix elements are the weights of the edges. For example, A[i][j]=0.1 indicates that node i and node j are "overlapping edges" with a weight of 0.1, which is suitable for the topological convolution operation of GNN.

[0061] Finally, the adjacency matrix is ​​transformed into graph structure data that can be directly read by a graph convolutional neural network. This graph structure data includes: a node feature matrix (X), an edge index matrix (edge_index), an edge attribute matrix (edge_attr), and additional information. The node feature matrix (X) has dimensions (N, F), where N is the total number of nodes (including color block nodes and reference vector nodes), and F is the feature dimension of a single node. The edge index matrix (edge_index) has dimensions (2, E), where E is the total number of edges, and each row stores the start and end node indices of the edge. The edge attribute matrix (edge_attr) has dimensions (E, 1), storing the weight / label encoding of each edge. The additional information is used to associate metadata such as the original land parcel ID and vector layer name, facilitating the traceability and matching of subsequent compliance analysis results with the verification report.

[0062] Another embodiment of this application also provides a land use planning verification device for performing the above-described method, the device comprising: The data collection module is used to acquire the planning vector map and color block planning map of the area to be verified; The color block recognition module is used to input the color block planning map into the convolutional neural network to extract color block features and obtain a multi-scale feature map; The multi-scale feature map is input into a pre-trained semantic segmentation network for color block segmentation, and the color block regions are output. The graph structure construction module is used to construct a topological structure graph of the area to be checked based on the color block area; The verification module is used to input the topology diagram and planning vector diagram into the graph neural network to analyze the topological relationships. By aggregating the characteristics of the target node itself and the characteristics of the neighboring nodes, the state prediction result of each target node is obtained. The state prediction result includes compliance, attribute violation, spatial violation or compound violation. The report generation module is used to integrate the state prediction results and generate a structured verification report based on the rule engine.

[0063] In some embodiments, the data collection module further includes a preprocessing unit, which is used to correct the planning vector map and color block planning map to a unified coordinate system using a spatial correction tool.

[0064] In some embodiments, the convolutional neural network backbone is based on ResNet50, and an SE attention module is inserted into ResNet50.

[0065] In some embodiments, the graph structure construction module is specifically used to: extract the geometric and attribute features of the color block region and establish initial topological relationships; Each color block area is set as a target node, and the control area is extracted from the planning vector map. Each control area is set as a reference vector node. Through spatial analysis and attribute matching, the node relationship between the target node and the reference vector node is quantified. Use spatial analysis tools to determine edge types and edge relationships, and define the relationships between edges in the graph based on the spatial relationships between target nodes and between target nodes and reference nodes; The initial topological relationships are corrected based on the node relationships and edge relationships to generate a topological structure graph for input to the graph neural network.

[0066] The specific limitations of the land use planning verification device provided in this embodiment can be found in the embodiments of the land use planning verification method described above, and will not be repeated here. Each module in the above-mentioned land use planning verification device can be implemented entirely or partially through software, hardware, or a combination thereof. Each module can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0067] This application provides a computer device that may include a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface communicates with external terminals via a network connection. When the computer program is executed by the processor, it causes the processor to perform the steps of the land use planning verification method as described in any of the above embodiments.

[0068] The working process, working details, and technical effects of the computer equipment provided in this embodiment can be found in the embodiment of the land use planning verification method described above, and will not be repeated here.

[0069] This application provides a computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the steps of the land use planning verification method as described in any of the above embodiments. The computer-readable storage medium refers to a data storage medium, which may include, but is not limited to, floppy disks, optical disks, hard disks, flash memory, USB flash drives, and / or memory sticks. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.

[0070] The working process, working details, and technical effects of the computer-readable storage medium provided in this embodiment can be found in the embodiment of the land use planning verification method described above, and will not be repeated here.

[0071] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0072] The technical features of the above embodiments can be combined in any way. For the sake of brevity, 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.

[0073] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the system described in this application can be divided into different functional units or modules to complete all or part of the functions described above.

Claims

1. A method for verifying land use planning, characterized in that, The method includes: Obtain the planning vector map and color block planning map of the area to be verified; The color block planning map is input into a convolutional neural network to extract color block features, resulting in a multi-scale feature map. The multi-scale feature map is input into a pre-trained semantic segmentation network for color block segmentation, and the color block regions are output. Based on the color block regions, construct a topological structure diagram of the region to be verified; The topology diagram and planning vector diagram are input into a graph neural network to analyze the topological relationships. By aggregating the characteristics of the target node itself and the characteristics of the neighboring nodes, the state prediction result of each target node is obtained. The state prediction result includes compliance, attribute violation, spatial violation or compound violation. The state prediction results are integrated, and a structured verification report is generated based on the rule engine.

2. The land use planning verification method according to claim 1, characterized in that, Obtaining the planning vector map and color block planning map of the area to be verified also includes: Use spatial correction tools to correct the planning vector map and color block planning map to a unified coordinate system.

3. The land use planning verification method according to claim 1, characterized in that, The convolutional neural network backbone uses ResNet50 as its base architecture, and an SE attention module is inserted into ResNet50.

4. The land use planning verification method according to claim 1, characterized in that, Based on the color block regions, a topological structure diagram of the region to be verified is constructed, including: Extract the geometric and attribute features of the color block regions and establish initial topological relationships; Each color block area is set as a target node, and the control area is extracted from the planning vector map. Each control area is set as a reference vector node. Through spatial analysis and attribute matching, the node relationship between the target node and the reference vector node is quantified. Use spatial analysis tools to determine edge types and edge relationships, and define the relationships between edges in the graph based on the spatial relationships between target nodes and between target nodes and reference nodes; The initial topological relationships are corrected based on the node relationships and edge relationships to generate a topological structure graph for input to the graph neural network.

5. A land use planning verification device, characterized in that, The apparatus for performing the method as described in any one of claims 1 to 7 comprises: The data collection module is used to acquire the planning vector map and color block planning map of the area to be verified; The color block recognition module is used to input the color block planning map into the convolutional neural network to extract color block features and obtain a multi-scale feature map; The multi-scale feature map is input into a pre-trained semantic segmentation network for color block segmentation, and the color block regions are output. The graph structure construction module is used to construct a topological structure graph of the area to be checked based on the color block area; The verification module is used to input the topology diagram and planning vector diagram into the graph neural network to analyze the topological relationships. By aggregating the characteristics of the target node itself and the characteristics of the neighboring nodes, the state prediction result of each target node is obtained. The state prediction result includes compliance, attribute violation, spatial violation or compound violation. The report generation module is used to integrate the state prediction results and generate a structured verification report based on the rule engine.

6. The land use planning verification device according to claim 5, characterized in that, The data collection module also includes a preprocessing unit, which is used to correct the planning vector map and color block planning map to a unified coordinate system using a spatial correction tool.

7. The land use planning verification device according to claim 5, characterized in that, The convolutional neural network backbone uses ResNet50 as its base architecture, and an SE attention module is inserted into ResNet50.

8. The land use planning verification device according to claim 5, characterized in that, The graph structure building module is specifically used for: Extract the geometric and attribute features of the color block regions and establish initial topological relationships; Each color block area is set as a target node, and the control area is extracted from the planning vector map. Each control area is set as a reference vector node. Through spatial analysis and attribute matching, the node relationship between the target node and the reference vector node is quantified. Use spatial analysis tools to determine edge types and edge relationships, and define the relationships between edges in the graph based on the spatial relationships between target nodes and between target nodes and reference nodes; The initial topological relationships are corrected based on the node relationships and edge relationships to generate a topological structure graph for input to the graph neural network.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the method according to any one of claims 1 to 4.

10. A computer device, characterized in that, The method includes 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 steps of the method according to any one of claims 1 to 4.

Images