Land space planning method and device based on big data, equipment and medium

Abstract

The invention relates to a land space planning method and device based on big data, equipment and a medium. The method comprises the following steps: acquiring territorial multi-source data, performing cloud removal processing on a remote sensing image, combining a digital elevation model and soil attribute data, constructing a spatial feature vector, calculating suitability scores of agriculture, town and ecology, and constructing a three-function score triple; calculating a conflict index according to the triple, generating and dividing a territorial space region based on a subject function tag, and recognizing a high-load region and a low-heat region in combination with a population thermodynamic diagram and infrastructure vector data; and identifying a to-be-migrated facility set in the high-load region, and generating a facility migration scheme and a territorial space planning vector layer through Nash equilibrium game optimization in the low-thermal region. According to the method, the precision and scientificity of territorial space planning are improved and the feasibility of a planning scheme is enhanced by integrating multi-source data and combining multi-objective optimization and a Nash equilibrium game to generate a facility migration scheme.

Description

Technical Field

[0001] This invention belongs to the field of computer technology, and in particular relates to methods, devices, equipment and media for land spatial planning based on big data. Background Technology

[0002] Territorial spatial planning is a core technical means to coordinate the allocation of various spatial resources, harmonize urban and rural development and ecological progress, and ensure the modernization of national spatial governance. It requires comprehensive consideration of the balance between three core functions: agricultural production, urban construction, and ecological protection, while also taking into account dynamic factors such as population distribution and infrastructure layout, in order to achieve the sustainable use of national land space. Currently, with the development of big data and spatial information technology, territorial spatial planning has gradually shifted from traditional qualitative analysis to quantitative modeling. However, existing technologies still have many shortcomings and cannot meet the needs of high-precision, multi-objective, and dynamic planning.

[0003] On the one hand, existing planning methods have limited ability to integrate multi-source land data in the data processing stage. They often rely on single-source data analysis or simple data splicing, failing to generate unified spatial feature vectors through spatial embedding and coding. This results in insufficient mining of the correlation between different types of data and loss of spatial feature information behind the data, directly affecting the accuracy of subsequent suitability assessments. On the other hand, in the regional delineation and facility optimization stages, existing planning methods often divide developed and undeveloped areas based on static land use status data, without combining functional labels, population heat maps, and infrastructure vector data to calculate regional heat values, making it difficult to accurately identify high-load and low-heat areas. Furthermore, in the formulation of facility relocation plans, these methods typically focus only on single-objective site selection based on the facility's own operating costs or spatial distance. They fail to identify the set of facilities to be relocated in high-load areas and then use Nash equilibrium game optimization to coordinate multiple objectives in low-heat areas. This fails to balance the interests of multiple stakeholders, including the government, developers, and residents, resulting in poor feasibility and adaptability of facility relocation plans. Summary of the Invention

[0004] Therefore, it is necessary to provide big data-based methods, devices, equipment, and media for land and space planning to address the aforementioned technical issues, aiming to improve the accuracy and scientific nature of land and space planning and enhance the feasibility and adaptability of planning schemes.

[0005] Firstly, this application provides a big data-based land spatial planning method, including:

[0006] Acquire multi-source land data including remote sensing images, digital elevation models, soil attribute data, population heat maps, and infrastructure vector data; perform cloud removal processing on the remote sensing images to generate cloud-removed optical images; and generate spatial feature vectors based on the cloud-removed optical images, digital elevation models, and soil attribute data through spatial embedding coding.

[0007] Based on spatial feature vectors, scores are calculated through three independent fully connected neural networks to obtain and construct a three-function score triplet based on agricultural suitability scores, urban suitability scores, and ecological suitability scores.

[0008] The conflict index is calculated based on the three-function score triplet. When the conflict index meets the preset conditions, the three-function score triplet is concatenated and fused with the spatial feature vector according to the dimension to generate the main function label based on the decision feature matrix and through the decision tree model.

[0009] Based on the main functional labels, the land space is divided into developed and undeveloped areas. Combined with population heat maps and infrastructure vector data, regional heat values ​​are calculated, and high-load areas and low-heat areas are identified through regional heat values.

[0010] In high-load areas, a set of facilities to be relocated is identified. Based on this set, a Nash equilibrium game optimization process is used in low-heat areas to generate a facility relocation plan. The main function labels and the facility relocation plan are then overlaid on the geographic information system to generate a land spatial planning vector layer.

[0011] In one embodiment, a spatial feature vector is generated based on de-clouded optical imagery, a digital elevation model, and soil property data through spatial embedding coding, including:

[0012] Pixel-level normalization is performed on the de-clouded optical images to generate standardized optical images;

[0013] Standardized optical images, digital elevation models, and soil property data are raster-vector coordinate aligned to generate a spatial registration dataset.

[0014] The spatial registration dataset is processed by a graph convolutional network to extract features and generate spatial feature vectors.

[0015] In one embodiment, a conflict index is calculated based on the three-function score triplet. When the conflict index meets a preset condition, the three-function score triplet is concatenated and fused with the spatial feature vector by dimension to generate a decision feature matrix. Then, based on this decision feature matrix, a main function label is generated using a decision tree model, including:

[0016] The conflict index is calculated based on the agricultural suitability score, urban suitability score, and ecological suitability score in the three-function score triplet.

[0017] When the conflict index is greater than the preset conflict threshold, it is determined that the conflict index meets the preset conditions. The three-function scoring triplet and the spatial feature vector are concatenated by dimension to generate an enhanced feature vector.

[0018] The enhanced feature vectors are processed by a convolutional neural network to generate a decision feature matrix;

[0019] The decision feature matrix is ​​input into the decision tree model for classification processing to generate main function labels.

[0020] In one embodiment, the land space is divided into developed and undeveloped areas based on the main functional labels, and regional heat values ​​are calculated by combining population heat maps and infrastructure vector data. High-load and low-heat areas are then identified based on these regional heat values, including:

[0021] Based on the main functional labels, the land space areas marked with urban functions are classified as developed areas, and the land space areas marked with agricultural and ecological functions are classified as undeveloped areas.

[0022] The mean normalized vegetation index of undeveloped areas is calculated based on remote sensing images. When the mean normalized vegetation index of undeveloped areas meets the preset mean threshold, the undeveloped areas are classified as major low-heat zones.

[0023] Within the developed areas, population density data is extracted based on population heat maps, and facility density data is extracted based on infrastructure vector data;

[0024] The regional thermal value is calculated based on population density data and facility density data;

[0025] When the regional thermal value is greater than the first thermal threshold, the corresponding developed area is identified as a high-load area;

[0026] When the regional thermal value is less than the second thermal threshold, the corresponding developed area is identified as a supplementary low thermal zone, and combined with the main low thermal zone, a low thermal zone is generated.

[0027] In one embodiment, a set of facilities to be relocated is identified in a high-load zone. Based on this set, a facility relocation plan is generated in a low-heat zone through Nash equilibrium game optimization, including:

[0028] Based on infrastructure vector data, the utilization rate data of each facility in the high-load area is extracted, and facilities whose utilization rate data exceeds the preset utilization rate threshold are selected to form a set of facilities to be relocated.

[0029] The protection coefficient was calculated based on the mean normalized vegetation index and the predicted GDP per unit area in the low heat zone.

[0030] A three-party utility function is constructed, which includes a government utility function, a developer utility function, and a resident utility function. The government utility function is constructed by the ratio of the actual resource usage to the resource carrying capacity in the high-load area. The developer utility function is constructed based on the predicted land value-added revenue and protection coefficient in the low-heat area. The resident utility function is constructed based on the difference in commuting time between the original location of the set of facilities to be relocated and the candidate locations in the low-heat area.

[0031] With the goal of maximizing the product of the government's utility function, the developer's utility function, and the resident's utility function, the Nash equilibrium solution is iteratively solved for the relocation locations of the set of facilities to be relocated within the low heat zone, generating a facility relocation scheme.

[0032] In one embodiment, the main function labels and facility relocation plans are overlaid onto a geographic information system to generate a land spatial planning vector layer, including:

[0033] The main functional labels are converted from raster to vector to generate a territorial spatial functional zoning surface layer;

[0034] Extract the coordinates of the new facilities from the facility relocation plan, perform coordinate transformation on the new facility coordinates, and generate a vector point layer;

[0035] The spatial inclusion relationship between each facility point in the vector point layer and the land space functional zoning surface layer is calculated using the spatial connection operator of the geographic information system.

[0036] Based on spatial inclusion relationships, attribute association processing is performed between the land spatial functional zoning surface layer and the vector point layer to generate a land spatial planning vector map layer.

[0037] In one embodiment, the conflict index is calculated using the following formula:

[0038]

[0039] in, As a conflict index, and These are the dispersion contribution coefficient and the ecological priority contribution coefficient, respectively, and satisfy the following conditions: , For suitability scoring, Corresponding agricultural suitability score , Corresponding town suitability score , Corresponding ecological suitability score , The average of the three suitability scores.

[0040] Secondly, this application also provides a land spatial planning device based on big data, including:

[0041] The multi-source data acquisition and integration module is used to acquire multi-source land data, including remote sensing images, digital elevation models, soil attribute data, population heat maps, and infrastructure vector data. It performs cloud removal processing on the remote sensing images to generate cloud-removed optical images, and generates spatial feature vectors based on the cloud-removed optical images, digital elevation models, and soil attribute data through spatial embedding coding.

[0042] The functional score calculation module is used to calculate scores based on spatial feature vectors through three independent fully connected neural networks, and to construct a three-functional score triplet based on agricultural suitability score, urban suitability score and ecological suitability score.

[0043] The conflict index assessment module is used to calculate the conflict index based on the three-function score triplet. When the conflict index meets the preset conditions, the three-function score triplet is concatenated and fused with the spatial feature vector according to the dimension to generate and generate the main function label based on the decision feature matrix through the decision tree model.

[0044] The regional division and load identification module is used to divide the national land space into developed and undeveloped areas based on the main function labels, and calculate the regional heat value by combining population heat map and infrastructure vector data, and identify high load areas and low heat areas through the regional heat value;

[0045] The facility relocation planning module is used to identify sets of facilities to be relocated in high-load areas. Based on these sets, a facility relocation plan is generated in low-heat areas through Nash equilibrium game optimization. The main functional labels and the facility relocation plan are then overlaid on the geographic information system to generate a land spatial planning vector layer.

[0046] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the first aspect.

[0047] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the first aspect.

[0048] The aforementioned big data-based land spatial planning methods, devices, equipment, and media first acquire multi-source land data and perform cloud removal processing on remote sensing images. Then, they combine this data with digital elevation models and soil attribute data to generate spatial feature vectors, addressing the shortcomings of existing methods such as insufficient integration of multi-source data and loss of spatial feature information. Secondly, based on the spatial feature vectors, three independent fully connected neural networks are used to construct three-function scoring triplets, avoiding the one-sidedness of single-function suitability assessments and improving the comprehensiveness and quantitative accuracy of the three core functions of land spatial planning. Furthermore, conflict indices are calculated based on the three-function scoring triplets to generate main function labels, solving the problem of traditional conflict determination relying on empirical qualitative analysis and enhancing the scientific rigor and accuracy of functional positioning in conflict areas. Finally, by combining the main function labels to divide areas and calculating heat values ​​to identify high and low load zones, a facility relocation scheme is generated based on Nash equilibrium game optimization and overlaid to generate a planning layer. This effectively solves the problems of static regional division, singular facility optimization, and imbalance of interests among multiple stakeholders, significantly improving the effectiveness of regional identification and the feasibility of facility relocation schemes. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0050] Figure 1 A flowchart of a land spatial planning method based on big data, provided as an exemplary embodiment of the present invention;

[0051] Figure 2 A flowchart of a method for generating main function tags is provided as an exemplary embodiment of the present invention;

[0052] Figure 3 This is a schematic diagram of a land spatial planning device based on big data, provided as an exemplary embodiment of the present invention. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0054] In one embodiment, such as Figure 1As shown, a land spatial planning method based on big data is provided. This embodiment illustrates the application of this method to a terminal. It is understood that this method can also be applied to a server, and further to a system including both a terminal and a server, and is implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0055] S101: Acquire multi-source land data including remote sensing images, digital elevation models, soil attribute data, population heat maps, and infrastructure vector data; perform cloud removal processing on the remote sensing images to generate cloud-removed optical images; and generate spatial feature vectors based on the cloud-removed optical images, digital elevation models, and soil attribute data through spatial embedding coding.

[0056] Specifically, territorial spatial planning needs to comprehensively reflect core elements such as landform, resource endowment, and ecological conditions. Therefore, remote sensing imagery, digital elevation models (DEMs), soil attribute data, population heat maps, and infrastructure vector data can be selected as multi-source inputs. Remote sensing imagery can provide information on land cover type and distribution; DEMs can reflect topographic features such as slope and elevation; soil attribute data is related to agricultural production and ecological carrying capacity; and population heat maps and infrastructure vector data correspond to the needs and current status of urban development. As an illustration, because remote sensing imagery is easily obscured by clouds due to atmospheric transmission, resulting in missing surface information, directly affecting the accuracy of subsequent feature extraction, it can first undergo cloud removal processing. Atmospheric interference can be eliminated through radiometric correction. For example, histogram matching methods can be used to replace pixel values ​​in cloud-covered areas with the average pixel values ​​of the same type of land cover in adjacent cloudless areas, generating cloud-removed optical images without information bias. Subsequently, a multi-source data fusion model can be constructed using deep learning frameworks such as TensorFlow or PyTorch to encode the de-clouded optical imagery, digital elevation models, and soil property data, generating spatial feature vectors. These vectors fully reflect the multi-dimensional information of the national land space, providing a feature foundation for subsequent analysis and modeling. This spatial encoding process achieves a unified representation of multi-source data while preserving pixel-level spatial context information through spatial correlation modeling. Compared to simple data stitching, this allows subsequent evaluation models to more accurately capture the synergistic features of topography, soil, and land cover.

[0057] S102: Based on spatial feature vectors, scores are calculated through three independent fully connected neural networks to obtain and construct a three-function score triplet based on agricultural suitability scores, urban suitability scores, and ecological suitability scores.

[0058] Specifically, the suitability factors for agriculture, urban areas, and ecological functions differ fundamentally. Agricultural suitability depends on resource conditions such as soil fertility and terrain slope; urban suitability is related to terrain flatness and spatial accessibility; and ecological suitability is associated with ecological elements such as vegetation cover and distance to water sources. Furthermore, fully connected neural networks excel at handling the nonlinear mapping relationship between structured feature vectors and target values. Therefore, employing three independent fully connected neural networks can achieve accurate modeling of each function's score, avoiding cross-factor interference. Illustratively, the three networks share the same input layer (to receive the generated spatial feature vectors), but the hidden layer parameters are optimized independently. For example, the first fully connected neural network uses the spatial feature vectors of historical agricultural planting areas and actual yield ratings as training data to learn the mapping law between spatial features and agricultural suitability; the second fully connected neural network uses the feature vectors of developed urban areas and development intensity as training data; and the third fully connected neural network uses the feature vectors of ecological protection zones and ecological quality indices as training data. The training process of these neural networks can use the Adam optimizer to minimize the mean squared error loss until the model's accuracy on the validation set stabilizes above a preset accuracy threshold (e.g., 90%), at which point training stops. During subsequent scoring, the spatial feature vector can be input into the three trained networks, which will output agricultural suitability scores, urban suitability scores, and ecological suitability scores ranging from [0,1], respectively. A score closer to 1 indicates higher suitability. By combining the three scores in the order of agriculture-urban-ecological, a three-function score triplet can be constructed. This triplet can intuitively reflect the suitability distribution of a region across the three functions, providing a clear quantitative basis for subsequent conflict identification.

[0059] S103: Calculate the conflict index based on the three-function score triplet. When the conflict index meets the preset conditions, the three-function score triplet and the spatial feature vector are concatenated and fused according to the dimension to generate the main function label based on the decision feature matrix and through the decision tree model.

[0060] Specifically, overlapping suitability scenarios are common in national land space, such as a region possessing potential for both agricultural and urban development. Therefore, a conflict index can quantify the degree of conflict; a higher index indicates more severe conflict between functions. Preset conditions can be established based on historical planning data and regional development goals. For example, a conflict threshold of 2.5 can be set for plains areas. When the calculated conflict index exceeds this value, the area is identified as a high-conflict zone, requiring the initiation of a refined decision-making process. When the conflict index meets the preset conditions, the three-function score triplet and spatial feature vector can be concatenated dimensionally to integrate original spatial features and suitability information. By comprehensively considering functional scores and spatial features, a decision feature matrix is ​​obtained. This matrix retains basic spatial features such as topography and soil while supplementing suitability quantification information, improving the accuracy of the decision-making model. Among these, the decision tree model, due to its advantages in processing structured data and providing highly interpretable output results, can be used to generate main function labels. Indicatively, this decision tree model can use areas with clearly defined functions in historical planning as samples, taking the decision feature matrix as input and the corresponding agricultural, urban, and ecological function labels as outputs. The model is trained until the node splitting gain stabilizes, resulting in a trained decision tree model. Inputting the decision feature matrix of the area to be evaluated into the trained decision tree model generates a unique primary function label.

[0061] S104: Based on the main functional labels, the national land space is divided into developed and undeveloped areas. Combined with population heat maps and infrastructure vector data, regional heat values ​​are calculated, and high-load areas and low-heat areas are identified through regional heat values.

[0062] Specifically, the main functional labels directly correspond to the core planning positioning of the national land space. For example, the urban functional label represents the established development pattern and can therefore be classified as developed areas. The agricultural and ecological functional labels correspond to the production and ecological spaces that need protection and can be classified as undeveloped areas. The load status of developed areas is determined by both population density and infrastructure service pressure. Therefore, the regional heat value can be calculated by combining population heat maps and infrastructure vector data. This value can comprehensively reflect the degree of matching between the region's development vitality, infrastructure support capacity, and main functional positioning. Based on the regional heat value, high-load areas and low-heat areas are identified. High-load areas are usually densely populated, under great infrastructure pressure, and have high development intensity, and may have problems such as resource shortages and environmental pressure. Low-heat areas may be sparsely populated, have insufficient infrastructure, and have great development potential. Accurately identifying these areas is of great significance for the optimized layout of national land space and the rational allocation of resources, and can provide a basis for subsequent facility relocation and coordinated regional development.

[0063] S105: Identify the set of facilities to be relocated in the high-load area, and based on the set of facilities to be relocated, generate a facility relocation plan by optimizing the plan through Nash equilibrium game in the low-heat area. Overlay the main function label and the facility relocation plan onto the geographic information system to generate a land spatial planning vector layer.

[0064] Specifically, identifying a set of facilities to be relocated within high-load areas can alleviate resource pressure and environmental burden, and optimize the regional functional layout. For example, by analyzing the distribution, operation, and impact of facilities in high-load areas, facilities that have a significant negative impact on regional development or are inconsistent with the main functional positioning can be identified as a set of facilities to be relocated. Based on this set of facilities to be relocated, a facility relocation plan can be generated in low-heat areas through Nash equilibrium game optimization. This Nash equilibrium game is a multi-agent decision optimization method that considers the interests and behavioral strategies of different stakeholders such as the government, developers, and residents, and finds the optimal facility relocation location and layout plan in low-heat areas. Through this game optimization process, the rational relocation of facilities and the coordinated development of regional functions can be achieved while satisfying the interests of all stakeholders. Geographic Information System (GIS) is a powerful spatial data analysis and visualization tool that can integrate and display various spatial data. By overlaying the main functional labels and facility relocation plans into the GIS as vector layers, the planning layout of the national land space and detailed information on facility relocation can be displayed intuitively. This provides clear guidance and visualization for planning implementation, facilitates planning management and supervision by relevant departments and personnel, and ensures the scientific and effective implementation of national land space planning.

[0065] The above method first acquires multi-source land data and performs cloud removal processing to generate spatial feature vectors, effectively addressing the limitations of data integration and the loss of spatial feature information in existing planning methods. Second, based on the spatial feature vectors, it calculates agricultural, urban, and ecological suitability scores, constructing a three-function score triplet, resolving functional conflicts in multi-objective planning and enhancing the scientific rigor of the plan. Furthermore, by calculating a conflict index and generating main function labels, the scientific nature of functional positioning is further improved. Moreover, by combining the main function labels to divide regions and calculating heat values ​​to identify high and low load zones, it solves the problems of static regional division and inaccurate load zone identification, improving the accuracy of regional identification. Finally, in the high and low load zones, it generates facility relocation schemes through Nash equilibrium game theory and overlays them to generate a planning layer, resolving the problems of singular facility optimization and imbalance of interests among multiple stakeholders, significantly improving the feasibility of the schemes.

[0066] In one embodiment, a spatial feature vector is generated based on de-clouded optical imagery, a digital elevation model, and soil property data through spatial embedding coding, including:

[0067] Pixel-level normalization is performed on the de-clouded optical images to generate standardized optical images;

[0068] Standardized optical images, digital elevation models, and soil property data are raster-vector coordinate aligned to generate a spatial registration dataset.

[0069] The spatial registration dataset is processed by a graph convolutional network to extract features and generate spatial feature vectors.

[0070] Specifically, the pixel values ​​of declouded images are affected by factors such as imaging equipment parameters and light intensity, resulting in inconsistent numerical ranges. For example, the pixel value range for the visible light band is 0-255, and for the near-infrared band it is 0-65535. Directly inputting these values ​​into the model can lead to an imbalance in feature weights, affecting the accuracy of subsequent coding. Therefore, a min-max normalization algorithm can be used. Taking a single declouded optical image as the processing unit, the maximum (Max) and minimum (Min) values ​​of all pixels in the image are first calculated. Then, a normalization formula is used to map all pixel values ​​to the [0,1] interval, generating a standardized optical image. This normalization process can eliminate dimensional differences, making the contribution of features from different bands in the image to subsequent coding more balanced. However, since the initial spatial references of the three types of data are different—standardized optical images usually use an image coordinate system, digital elevation models use a geographic coordinate system, and soil attribute data are mostly stored as vector area features and associated with sampling point coordinates—direct overlay can lead to spatial misalignment, resulting in a mismatch between features and geographic locations. Therefore, all data can first be uniformly converted to the CGCS2000 coordinate system. Bilinear interpolation can then be used to resample the digital elevation model to the same spatial resolution as the standardized optical imagery, such as 10m × 10m. For soil attribute data, isometric features can be converted to raster data through vector rasterization. The raster cell value is assigned the mean of the soil attribute within that cell, such as soil pH. This ultimately generates a raster dataset with one-to-one spatial correspondence, i.e., a spatial registration dataset. This process ensures that the three types of data are correlated under the same spatial reference, providing a spatial consistency basis for subsequent fusion coding and avoiding semantic confusion caused by coordinate deviations.

[0071] Finally, traditional convolutional networks can only capture local pixel relationships, while spatial features such as topographic slope and soil fertility exhibit global relationships. Graph convolutional networks can model global spatial dependencies through graph structures. Therefore, graph convolutional networks can be used to extract features from spatial registration datasets. For example, each raster cell in the spatial registration dataset can be treated as a node in a graph, with node features consisting of a feature vector composed of the standardized optical image pixel value, digital elevation value, and soil attribute value corresponding to that cell. Based on Euclidean distance, each node is connected to its eight neighboring nodes to form edges, constructing a spatial graph structure. By aggregating the features of each node and its neighbors, the graph convolutional network learns the spatial feature relationships. After three layers of convolution and activation functions (such as ReLU), the high-dimensional graph features can be compressed into a fixed-dimensional vector, such as a 128-dimensional vector, i.e., a spatial feature vector. This vector reflects both the local attributes of a single cell and the global features at the regional scale, providing comprehensive spatial semantic support for subsequent suitability scoring.

[0072] In one embodiment, such as Figure 2 As shown, a conflict index is calculated based on the three-function score triplet. When the conflict index meets preset conditions, the three-function score triplet is concatenated and fused with the spatial feature vector according to its dimensions to generate a decision feature matrix. Based on this, a decision tree model is used to generate the main function labels, including:

[0073] S201: Calculate the conflict index based on the agricultural suitability score, urban suitability score, and ecological suitability score in the three-function score triplet;

[0074] S202: When the conflict index is greater than the preset conflict threshold, it is determined that the conflict index meets the preset conditions. The three-function scoring triplet and the spatial feature vector are concatenated by dimension to generate an enhanced feature vector.

[0075] S203: The enhanced feature vector is processed by a convolutional neural network to generate a decision feature matrix;

[0076] S204: Input the decision feature matrix into the decision tree model for classification processing to generate main function labels.

[0077] Specifically, due to the competition among the three major functions of agriculture, urban areas, and ecology in national land spatial planning—that is, when a region exhibits high suitability in two or all three functions—it is difficult to determine the dominant function based solely on a single score. Therefore, this embodiment quantifies this competitive relationship through a conflict index. The formula for calculating this conflict index is illustrated below:

[0078]

[0079] in, As a conflict index, and These are the dispersion contribution coefficient and the ecological priority contribution coefficient, respectively, and satisfy the following conditions: , For suitability scoring, Corresponding agricultural suitability score , Corresponding town suitability score , Corresponding ecological suitability score , The average of the three suitability scores.

[0080] In the above formula, the agricultural suitability score is calculated first. Urban suitability score and ecological suitability score The average of the three factors reflects the overall suitability level of the region. Then, the coefficient of variation is calculated. The formula quantifies the dispersion of the three scores; the higher the dispersion, the more ambiguous the functional positioning and the more significant the basis for conflict. Furthermore, the formula incorporates an ecological priority factor. When the sum of urban and agricultural suitability exceeds ecological suitability, this item is positive, thus reflecting the conflict between urban and agricultural sectors on ecological space; otherwise, it is zero. Finally, through dual-dimensional weighting, it captures both the balance of functional suitability and the priority of ecological protection, improving the accuracy of conflict characterization in complex areas compared to the traditional single-dimensional conflict index.

[0081] Specifically, when the conflict index If the value exceeds a preset conflict threshold (e.g., 0.8 for ecologically sensitive areas and 1.0 for urban core areas), it can be determined that the preset conditions are met. In this case, since the spatial feature vector (e.g., 128-dimensional) only reflects basic spatial attributes such as topography and soil, while the three-function score triplet (3-dimensional) is a quantitative result of functional suitability, an enhanced feature vector can be generated by concatenating the three-function score triplet with the spatial feature vector according to their dimensions. This can eliminate the information bias of a single feature. For example, if the spatial feature vector is 128-dimensional and the three-function score triplet is 3-dimensional, the enhanced feature vector is 131-dimensional, with each dimension corresponding to the numerical value of the original feature or score, ensuring the integrity of the feature semantics.

[0082] Subsequently, a convolutional neural network (CNN) can be used to perform feature transformation on the enhanced feature vectors, mapping the high-dimensional linear features into a two-dimensional decision feature matrix rich in non-linear correlations, thus solving the feature redundancy problem caused by dimensionality stacking. The CNN can adopt an "input layer-convolutional layer-pooling layer-output layer" structure. The input layer receives the 131-dimensional enhanced feature vectors and reshapes them into a 1×131 one-dimensional matrix. The convolutional layer uses 32 1×3 convolutional kernels, calculating local feature correlations through a sliding window to generate a 32×129 feature map. The pooling layer uses 2×2 max pooling to compress the feature map to 16×65, and finally, global average pooling can compress the feature map into a 64×64 decision feature matrix. This feature transformation process, through convolutional operations, can capture implicit dependencies between features, further improving the discriminative power of feature representation.

[0083] Specifically, during the training of the decision tree model, areas with clearly defined functions in historical planning are used as samples. Elements of the decision feature matrix are used as input features, and agricultural / urban / ecological functions are used as output labels. The optimal splitting node, such as a threshold for a matrix element, is selected using the information gain ratio to progressively construct a classification tree (the maximum depth can be set to 10 to avoid overfitting). During classification, the decision feature matrix of the area to be evaluated can be judged along the tree path, such as "matrix element value > 0.6 → left subtree, otherwise right subtree," ultimately outputting a unique primary function label. Compared to traditional expert judgment, this process further improves the classification accuracy of function labels, and each decision node corresponds to specific feature criteria, enhancing the scientific rigor and interpretability of the functional positioning of conflict areas.

[0084] In one embodiment, the national land space is divided into developed and undeveloped areas based on the main functional labels, and regional heat values ​​are calculated by combining population heat maps and infrastructure vector data. High-load and low-heat areas are then identified based on these regional heat values, including:

[0085] Based on the main functional labels, the land space areas marked with urban functions are classified as developed areas, and the land space areas marked with agricultural and ecological functions are classified as undeveloped areas.

[0086] The mean normalized vegetation index of undeveloped areas is calculated based on remote sensing images. When the mean normalized vegetation index of undeveloped areas meets the preset mean threshold, the undeveloped areas are classified as major low-heat zones.

[0087] Within the developed areas, population density data is extracted based on population heat maps, and facility density data is extracted based on infrastructure vector data;

[0088] The regional thermal value is calculated based on population density data and facility density data;

[0089] When the regional thermal value is greater than the first thermal threshold, the corresponding developed area is identified as a high-load area;

[0090] When the regional thermal value is less than the second thermal threshold, the corresponding developed area is identified as a supplementary low thermal zone, and combined with the main low thermal zone, a low thermal zone is generated.

[0091] Specifically, based on the main functional labels, land space raster units marked "urban function" can be classified as developed areas, while raster units marked "agricultural function" and "ecological function" can be classified as undeveloped areas. Then, the mean normalized vegetation index (NDI) of the undeveloped areas is calculated, and major low-heat areas are selected. The NDI is a core indicator reflecting vegetation cover and ecological quality; a higher value indicates denser vegetation, stronger ecological sensitivity, and lower development potential, thus serving as a quantitative basis for the "low-heat" characteristic of undeveloped areas. For example, the near-infrared and red band reflectance of all pixels in the undeveloped area can be extracted from de-clouded optical imagery, and the NDI value can be calculated pixel-by-pixel to obtain the regional mean. The preset mean threshold can be set to 0.6 for ecologically sensitive areas and 0.4 for general agricultural areas. When the mean normalized vegetation index of an undeveloped area is greater than the threshold, it can be determined that the area has high ecological quality and strong development constraints, and is classified as a major low-heat zone as a core candidate area for facility relocation, ensuring the ecological compatibility of the relocation plan.

[0092] Furthermore, within developed areas, population density and facility density are key indicators reflecting the region's load status. Population density reflects the intensity of human activity's consumption of resources, while facility density reflects the service pressure on infrastructure. Illustratively, the population heat map uses raster data, allowing the extraction of population density values ​​for each raster cell within the developed area. Infrastructure vector data consists of point features, and the number of facilities within each raster cell can be statistically analyzed through spatial overlay to calculate facility density values. Subsequently, based on the population density and facility density values, a weighted linear combination is used to calculate the regional heat map value. The weight allocation can be based on the logic that "population is the main consumer of resources, and facilities are the carriers of load," ensuring that the heat map value simultaneously reflects both human demand and the supply pressure of facilities.

[0093] Furthermore, a first and second thermal threshold can be calculated by combining regional resource carrying capacity. For example, based on the carrying capacity limits of water and land resources, historical data can be used to fit the thermal values ​​of areas that have experienced resource overload. The threshold for high-load areas can be set to 1200, meaning that when the thermal value is >1200, the regional resource consumption is close to or exceeds the carrying capacity limit. The threshold for supplementary low-heat areas can be set to 300, meaning that when the thermal value is <300, the region has a low level of development and high resource redundancy. When the thermal value of a developed area is greater than the first thermal threshold, it can be identified as a high-load area. When the thermal value is less than the second thermal threshold, it is identified as a supplementary low-heat area, which, together with the main low-heat areas, constitutes a set of low-heat areas. This provides sufficient space for facility relocation, ensuring that the relocation plan has both ecological protection limits and development potential.

[0094] In one embodiment, a set of facilities to be relocated is identified in a high-load area, and based on this set, a facility relocation plan is generated in a low-heat area through Nash equilibrium game optimization, including:

[0095] Based on infrastructure vector data, the utilization rate data of each facility in the high-load area is extracted, and facilities whose utilization rate data exceeds the preset utilization rate threshold are selected to form a set of facilities to be relocated.

[0096] The protection coefficient was calculated based on the mean normalized vegetation index and the predicted GDP per unit area in the low heat zone.

[0097] A three-party utility function is constructed, which includes a government utility function, a developer utility function, and a resident utility function. The government utility function is constructed by the ratio of the actual resource usage to the resource carrying capacity in the high-load area. The developer utility function is constructed based on the predicted land value-added revenue and protection coefficient in the low-heat area. The resident utility function is constructed based on the difference in commuting time between the original location of the set of facilities to be relocated and the candidate locations in the low-heat area.

[0098] With the goal of maximizing the product of the government's utility function, the developer's utility function, and the resident's utility function, the Nash equilibrium solution is iteratively solved for the relocation locations of the set of facilities to be relocated within the low heat zone, generating a facility relocation scheme.

[0099] Specifically, the resource pressure in high-load areas stems from the imbalance between facility service capacity and demand. Utilization rate directly reflects the actual carrying capacity of facilities, and facilities with utilization rates exceeding thresholds are the core cause of excessive regional load. Therefore, relocating such facilities can maximize the relief of load pressure. Illustratively, operational data for each facility in the high-load area can be extracted from the attribute table of infrastructure vector data. For example, for public service facilities, the average daily service visits and designed service capacity can be extracted, and the utilization rate can be calculated using the ratio of these two. For production facilities, the average daily energy consumption and rated energy consumption can be extracted, and the utilization rate can be calculated as: Utilization Rate = Average Daily Energy Consumption / Rated Energy Consumption × 100%. Preset utilization rate thresholds can be set differently based on facility type. For public service facilities, which are directly related to people's livelihoods, the threshold can be set to 85% to avoid insufficient service capacity leading to a decline in experience. For production facilities, the threshold can be set to 90% to balance production efficiency and resource consumption. Subsequently, facilities with utilization rates exceeding the corresponding thresholds can be filtered out and organized into a set of facilities to be relocated, in the format of facility type - original location coordinates - designed capacity.

[0100] Furthermore, while low-heat zones have lower loads, some areas, such as woodlands and water source protection areas, are ecologically sensitive and require protection coefficients to limit over-development and prevent ecological damage from facility relocation. This can be achieved by first calculating the mean normalized vegetation index (NDI) of the low-heat zone; a higher value indicates higher vegetation cover and better ecological quality. The GDP per unit area can be predicted using the ARIMA time series model. Using historical GDP per unit area data for the administrative unit where the low-heat zone is located over a preset period as the training set, the average value for the next three years (facility operation cycle) is predicted as the GDP per unit area. A higher value indicates greater economic development potential for the region. Subsequently, the protection coefficient can be obtained by calculating the ratio of this average value to the predicted GDP per unit area. A larger value indicates stronger ecological constraints and requires more cautious development, providing a quantitative basis for ecological protection demands in subsequent negotiations.

[0101] Specifically, the utility functions for these three parties need to match the core demands of the government, developers, and residents respectively. The government, as the main body of spatial governance, aims to alleviate resource pressure in high-load areas. Therefore, its utility function is constructed based on the ratio of actual resource usage to resource carrying capacity. Actual resource usage refers to the average daily consumption of core resources such as water and electricity, while resource carrying capacity is the upper limit of resource supply in the regional plan. A negative value indicates a smaller ratio (lower load) and higher utility. Developers, as the main investors, aim to balance development revenue with ecological costs. Therefore, their utility function is constructed based on the predicted land appreciation revenue and the protection coefficient. The formula can be: = "Predicted Land Appreciation Gain - λ × Protection Coefficient", where the predicted land appreciation gain is calculated by the difference between the benchmark land price in low-heat zones and the expected land price after facility relocation, and λ is the ecological cost coefficient (calibrated using ecological compensation data from historical development projects), thus achieving a balance between revenue and ecological costs. Residents, as users of the facilities, primarily demand convenient commuting; therefore, the resident utility function is constructed based on the "commuting time difference," with the formula: ,in The average commute time from the facility's original location to the residential area (calculated using road network data and the shortest path algorithm). The average commuting time from candidate locations in the low-thermal-energy zone to the same cluster area is represented by a negative value, indicating that the smaller the commuting time difference (higher convenience) and the higher the utility. A gradient ascent method can then be used. First, ecologically sensitive units with a protection coefficient greater than 0.8 are eliminated within the low-thermal-energy zone to obtain candidate relocation areas. Each facility in the set of facilities to be relocated is used as the optimization object, and the candidate location is initialized as a random point within the candidate area. The three-way utility product of the current location is calculated, and the location is iteratively adjusted in steps of 0.01. The utility product is recalculated after each adjustment until the change in the product is less than a preset threshold after five consecutive iterations. This location is the Nash equilibrium solution for that facility. By repeating this process for all facilities to be relocated, a facility relocation scheme consisting of facility number, optimal relocation coordinates, and corresponding ecological compensation amount can be finally formed.

[0102] In one embodiment, the main function labels and facility relocation plans are overlaid onto a geographic information system to generate a land spatial planning vector layer, including:

[0103] The main functional labels are converted from raster to vector to generate a territorial spatial functional zoning surface layer;

[0104] Extract the coordinates of the new facilities from the facility relocation plan, perform coordinate transformation on the new facility coordinates, and generate a vector point layer;

[0105] The spatial inclusion relationship between each facility point in the vector point layer and the land space functional zoning surface layer is calculated using the spatial connection operator of the geographic information system.

[0106] Based on spatial inclusion relationships, attribute association processing is performed between the land spatial functional zoning surface layer and the vector point layer to generate a land spatial planning vector map layer.

[0107] Specifically, the main functional labels are initially stored in raster form (e.g., 100m×100m pixel units, each unit marked with a unique functional label). While this can accurately represent spatial distribution, the boundaries are blurry and it is not convenient for regional attribute statistics. In contrast, vector polygon layers precisely define the area range through polygon boundaries, supporting structured storage and spatial analysis of attribute information. Therefore, this embodiment adopts a raster-to-vector algorithm based on edge detection. First, the raster data is binarized (units with the same functional label are assigned a value of 1, and others are assigned 0). The Canny operator is used to extract the boundary pixels of the functional areas. Then, the Douglas-Puk algorithm is used to thin out the boundary pixels, removing redundant points to simplify the polygon structure, while retaining key inflection points to ensure boundary accuracy. Finally, adjacent polygons with the same functional label are merged, thereby generating a national spatial functional zoning polygon layer with functional label-area-boundary coordinates as attributes.

[0108] Furthermore, since the coordinates of the new facilities in the facility relocation plan may be based on a local coordinate system (such as a construction coordinate system), while the land spatial functional zoning surface layer uses the national geodetic coordinate system (such as CGCS2000), direct overlay will result in spatial misalignment, leading to incorrect association between the facilities and their respective functional zones. Therefore, this embodiment first extracts the three-dimensional coordinates of the new facilities from the attribute table of the facility relocation plan, and then determines the coordinate transformation parameters using the control point matching method. That is, by selecting more than three control points that exist simultaneously in both the local coordinate system and the CGCS2000 coordinate system (such as landmark buildings with known latitude and longitude), a seven-parameter transformation model (including three translation parameters, three rotation parameters, and one scale parameter) is calculated. By substituting the coordinates of the new facilities into the transformation model, the coordinates in the CGCS2000 coordinate system can be obtained. Finally, the transformed coordinates are stored in the form of point features, and attributes such as facility type, design capacity, and pre-relocation location are added to generate a vector point layer. Through this transformation process, the accurate spatial correspondence between the facility points and the functional zoning surface can be ensured.

[0109] Specifically, the spatial connectivity operator employs a point-surface inclusion analysis algorithm. For each facility point (x, y) in the vector point layer, it traverses each polygon in the land space functional zoning surface layer. A ray casting method is used to determine if a point is inside a polygon (radiating rays from the point in any direction and counting the number of intersections with the polygon boundary; an odd number indicates the point is inside, an even number indicates outside). For points located on polygon boundaries (such as those falling on functional zone boundaries), their distance from the polygon vertices is compared to assign them to a larger functional zone. After calculation, a "functional zone ID" attribute is added to each facility point, recording its functional zone. Since the functional zoning surface layer contains the core functional positioning of the land space, and the vector point layer contains the specific layout information of the facilities, the fusion of their attributes forms a complete planning element dataset. Therefore, this embodiment can use "functional zone ID" as the key to connect the attribute table of the land space functional zoning surface layer (including functional labels, area, control requirements, etc.) with the attribute table of the vector point layer (including facility type, design capacity, etc.), generating a composite layer containing both surface and point elements. Furthermore, the symbolic rendering function of the Geographic Information System (GIS) can be used to assign differentiated colors to different functional zones, such as light gray for urban functional zones, light green for agricultural functional zones, and dark green for ecological functional zones. Different symbols can also be used for different types of facilities, such as a cross symbol for hospitals and a book symbol for schools. Ultimately, this results in a land spatial planning vector layer that combines spatial location accuracy with complete attribute information. This layer can be directly used for the visualization of planning schemes, spatial conflict detection, and implementation progress tracking, providing intuitive and accurate technical support for land spatial governance.

[0110] Based on the same inventive concept, such as Figure 3 As shown, this application also provides a big data-based land spatial planning device 300 for implementing the big data-based land spatial planning method described above. The solution provided by this device is similar to the implementation scheme described in the above method; therefore, the specific limitations of one or more embodiments of the big data-based land spatial planning device provided below can be found in the limitations of the various method embodiments above, and will not be repeated here. The device includes:

[0111] The multi-source data acquisition and integration module 301 is used to acquire multi-source land data including remote sensing images, digital elevation models, soil attribute data, population heat maps and infrastructure vector data, perform cloud removal processing on the remote sensing images to generate cloud-removed optical images, and generate spatial feature vectors based on the cloud-removed optical images, digital elevation models and soil attribute data through spatial embedding coding.

[0112] The functional score calculation module 302 is used to calculate scores based on spatial feature vectors through three independent fully connected neural networks, and to construct a three-functional score triplet based on agricultural suitability score, urban suitability score and ecological suitability score.

[0113] The conflict index assessment module 303 is used to calculate the conflict index based on the three-function score triplet. When the conflict index meets the preset conditions, the three-function score triplet and the spatial feature vector are spliced ​​and fused according to the dimension to generate and generate the main function label based on the decision feature matrix through the decision tree model.

[0114] The regional division and load identification module 304 is used to divide the national land space into developed and undeveloped areas based on the main function labels, and calculate the regional heat value by combining the population heat map and infrastructure vector data, and identify high load areas and low heat areas through the regional heat value;

[0115] The facility relocation planning module 305 is used to identify a set of facilities to be relocated in high-load areas. Based on the set of facilities to be relocated, it generates a facility relocation plan through Nash equilibrium game optimization in low-heat areas. The main functional labels and the facility relocation plan are overlaid on the geographic information system to generate a land spatial planning vector layer.

[0116] In one exemplary embodiment, the present invention also provides a computer device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the big data-based land spatial planning method of this application. A multi-core processor is preferred to improve the system's parallel processing capability. The memory provides sufficient temporary storage space to support program execution and data processing. The memory capacity should be large enough to accommodate large amounts of data and computational tasks.

[0117] In one exemplary embodiment, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the big data-based land spatial planning method of this application. The computer-readable storage medium may include: a read-only memory, a random access memory (RAM), a solid-state drive (SSD), or an optical disc, etc.

[0118] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A land spatial planning method based on big data, characterized in that, The method includes: Acquire multi-source land data including remote sensing images, digital elevation models, soil attribute data, population heat maps, and infrastructure vector data; perform cloud removal processing on the remote sensing images to generate cloud-removed optical images; and generate spatial feature vectors based on the cloud-removed optical images, the digital elevation models, and the soil attribute data through spatial embedding coding. Based on the spatial feature vector, scores are calculated using three independent fully connected neural networks to obtain and construct a three-function score triplet based on agricultural suitability scores, urban suitability scores, and ecological suitability scores. The conflict index is calculated based on the three-function scoring triplet. When the conflict index meets the preset conditions, the three-function scoring triplet is concatenated and fused with the spatial feature vector according to the dimension to generate and generate the main function label based on the decision feature matrix through the decision tree model. Based on the main functional labels, the land space is divided into developed and undeveloped areas. Combined with the population heat map and the infrastructure vector data, the regional heat value is calculated, and high-load areas and low-heat areas are identified through the regional heat value. In the high-load area, a set of facilities to be relocated is identified. Based on the set of facilities to be relocated, a facility relocation plan is generated in the low-heat area through Nash equilibrium game optimization. The main function label and the facility relocation plan are overlaid on the geographic information system to generate a land spatial planning vector layer.

2. The method according to claim 1, characterized in that, The process of generating spatial feature vectors through spatial embedding coding based on the de-clouded optical imagery, the digital elevation model, and the soil attribute data includes: The de-clouded optical image is subjected to pixel-level normalization to generate a standardized optical image; The standardized optical image, the digital elevation model, and the soil property data are subjected to raster-vector coordinate alignment processing to generate a spatial registration dataset. The spatial registration dataset is processed by a graph convolutional network to extract features and generate the spatial feature vector.

3. The method according to claim 1, characterized in that, The process involves calculating a conflict index based on the three-function score triplet. When the conflict index meets a preset condition, the three-function score triplet is concatenated and fused with the spatial feature vector along its dimensions to generate a decision feature matrix. Then, based on this decision feature matrix, a main function label is generated using a decision tree model. This includes: The conflict index is calculated based on the agricultural suitability score, urban suitability score, and ecological suitability score in the three-function score triplet. When the conflict index is greater than the preset conflict threshold, it is determined that the conflict index meets the preset condition. The three-function scoring triplet and the spatial feature vector are concatenated by dimension to generate an enhanced feature vector. The enhanced feature vector is processed by a convolutional neural network to generate the decision feature matrix; The decision feature matrix is ​​input into the decision tree model for classification processing to generate main function labels.

4. The method according to claim 1, characterized in that, The process of dividing the national land space into developed and undeveloped areas based on the main functional labels, and calculating regional heat values ​​by combining the population heat map and the infrastructure vector data, and identifying high-load and low-heat areas through the regional heat values, includes: Based on the main functional labels, the land space areas marked with urban functions are classified as developed areas, and the land space areas marked with agricultural and ecological functions are classified as undeveloped areas. The mean normalized vegetation index of the undeveloped area is calculated based on the remote sensing image. When the mean normalized vegetation index of the undeveloped area meets the preset mean threshold, the undeveloped area is divided into the main low heat zone. Within the developed area, population density data is extracted based on the population heat map, and facility density data is extracted based on the infrastructure vector data; Calculate the regional thermal value based on the population density data and the facility density data; When the thermal value of the area is greater than the first thermal threshold, the corresponding developed area is identified as the high-load area; When the thermal value of the area is less than the second thermal threshold, the corresponding developed area is identified as a supplementary low thermal zone, and combined with the main low thermal zone, the low thermal zone is generated.

5. The method according to claim 1, characterized in that, The process of identifying a set of facilities to be relocated within the high-load zone, and generating a facility relocation plan based on this set of facilities within the low-heat zone through Nash equilibrium game optimization, includes: Based on the infrastructure vector data, the utilization rate data of each facility in the high-load area is extracted, and facilities whose utilization rate data exceeds a preset utilization rate threshold are selected to form the set of facilities to be relocated. The protection coefficient is calculated based on the mean normalized vegetation index and the predicted GDP per unit area of ​​the low-heat zone. A three-party utility function is constructed, comprising a government utility function, a developer utility function, and a resident utility function. The government utility function is constructed based on the ratio of actual resource usage to resource carrying capacity in the high-load area. The developer utility function is constructed based on the predicted land appreciation revenue and the protection coefficient in the low-heat area. The resident utility function is constructed based on the difference in commuting time between the original location of the set of facilities to be relocated and the candidate locations in the low-heat area. With the objective of maximizing the product of the government utility function, the developer utility function, and the resident utility function, the relocation locations of the set of facilities to be relocated within the low-heat zone are iteratively solved to obtain the Nash equilibrium solution, thereby generating the facility relocation scheme.

6. The method according to claim 1, characterized in that, The step of overlaying the main functional tags and the facility relocation plan onto the geographic information system to generate a land spatial planning vector layer includes: The main functional labels are processed by raster-to-vector conversion to generate a land space functional zoning surface layer; Extract the coordinates of the new facility from the facility relocation plan, perform coordinate transformation on the new facility coordinates, and generate a vector point layer; The spatial inclusion relationship between each facility point in the vector point layer and the land spatial functional zoning surface layer is calculated using the spatial connection operator of the geographic information system. Based on the spatial inclusion relationship, the land spatial functional zoning surface layer and the vector point layer are subjected to attribute association processing to generate the land spatial planning vector map layer.

7. The method according to claim 3, characterized in that, The conflict index is calculated using the following formula: in, The conflict index is... and These are the dispersion contribution coefficient and the ecological priority contribution coefficient, respectively, and satisfy the following conditions: , For suitability scoring, Corresponding to the agricultural suitability score , Corresponding to the town suitability score , Corresponding to the ecological suitability score , The mean of the three suitability scores.

8. A land spatial planning device based on big data, characterized in that, The device includes: The multi-source data acquisition and integration module is used to acquire multi-source land data including remote sensing images, digital elevation models, soil attribute data, population heat maps and infrastructure vector data. It performs cloud removal processing on the remote sensing images to generate cloud-removed optical images, and generates spatial feature vectors based on the cloud-removed optical images, the digital elevation models and the soil attribute data through spatial embedding coding. The functional score calculation module is used to calculate scores based on the spatial feature vector through three independent fully connected neural networks, and to construct a three-functional score triplet based on the agricultural suitability score, urban suitability score and ecological suitability score. The conflict index assessment module is used to calculate the conflict index based on the three-function scoring triplet. When the conflict index meets the preset conditions, the three-function scoring triplet is concatenated and fused with the spatial feature vector according to the dimension to generate and generate the main function label based on the decision feature matrix through the decision tree model. The regional division and load identification module is used to divide the national land space into developed and undeveloped areas based on the main functional labels, and calculate the regional heat value by combining the population heat map and the infrastructure vector data, and identify high load areas and low heat areas through the regional heat value; The facility relocation planning module is used to identify a set of facilities to be relocated in the high-load area, and based on the set of facilities to be relocated, generate a facility relocation plan through Nash equilibrium game optimization in the low-heat area. The main functional labels and the facility relocation plan are overlaid on the geographic information system to generate a land spatial planning vector layer.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the 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, it implements the steps of the method according to any one of claims 1 to 7.

Images