A land use planning optimization method and system based on spatiotemporal big data

Abstract

This invention relates to the field of land planning technology, specifically to a land use planning optimization method and system based on spatiotemporal big data. A land use planning optimization system based on spatiotemporal big data includes: a historical land use data sequence acquisition module, a change characteristic calculation module, and an iterative optimization module. This invention acquires multi-temporal land use remote sensing raster map sequences of the land to be planned, constructs a pixel-level land use change dataset, comprehensively characterizes the land use evolution trends and change sensitivity of each location at different historical stages, and ensures that the model can identify the spatial distribution characteristics of stable areas and high-frequency change areas. Based on land type change statistics, a global transition matrix and trend probability vector are constructed to achieve quantitative modeling of land use evolution patterns, and this is used for trend-guided construction of the initial population of a genetic algorithm, improving the directionality and convergence efficiency of the optimization.

Description

Technical Field

[0001] This invention relates to the field of land planning technology, specifically to a land use planning optimization method and system based on spatiotemporal big data. Background Technology

[0002] With the acceleration of urbanization and increasing pressure on resources and the environment, land use planning has become a key means to achieve efficient allocation of national land space, ecological protection, and sustainable development. Traditional land use planning methods mainly rely on static data and expert experience for land allocation, often employing objective functions and heuristic algorithms for optimization. Typical methods include models based on cellular automata, genetic algorithms, and multi-objective programming. While these methods have improved the rationality of land use layout and the ability to balance multiple objectives to some extent, they have significant shortcomings in modeling and utilizing the dynamic evolution trends and historical spatiotemporal distribution characteristics of land use. Summary of the Invention

[0003] This invention acquires multi-temporal land use remote sensing raster image sequences of land to be planned, constructs a pixel-level land use change dataset, comprehensively characterizes the land use evolution trends and change sensitivity of each location at different historical stages, and ensures that the model can identify the spatial distribution characteristics of stable areas and high-frequency change areas. Based on land type transformation statistics, a global transition matrix and trend probability vector are constructed to achieve quantitative modeling of land use evolution patterns, which are then used for trend-guided construction of the initial population of the genetic algorithm, improving the directionality and convergence efficiency of optimization. A fitness function is constructed by combining multiple objectives such as ecosystem service value, economic output capacity, spatial structure rationality, and trend consistency to achieve a comprehensive evaluation of ecological protection, economic development, and spatial coordination. During the evolution process, a patch identification mechanism and neighborhood spatial rule control strategy are introduced to maintain the integrity and coherence of the land pattern, and local restoration operations are used to suppress fragmentation and leapfrog development, thereby improving the practical feasibility and spatial rationality of the optimization results.

[0004] This invention provides a land use planning optimization method based on spatiotemporal big data, comprising:

[0005] Obtain the historical land use data sequence of the land to be planned. The historical land use data sequence includes historical land use raster maps arranged in chronological order, and each cell in the historical land use raster map stores the land type of the corresponding location of the land to be planned at the corresponding time point in the historical land use raster map.

[0006] The number of land type changes in the historical land use data sequence is counted to construct a global land transfer matrix. For each cell of the land to be planned, the following steps are performed: the frequency of each land type appearing in the cell in the historical land use data sequence is counted, the land change probability vector corresponding to the cell is constructed, the number of land type changes in the cell in the historical land use data sequence is counted, the change frequency value corresponding to the cell is calculated, the last historical land use raster in the historical land use data sequence is selected, the land type of the cell in the selected historical land use raster is determined, and the row vector in the global land transfer matrix is ​​selected as the land trend probability vector based on the determined land type.

[0007] A population set is constructed from several land use planning simulation individuals. These individuals are constructed based on land change probability vectors, land trend probability vectors, and change frequency values. The storage format of the land use planning simulation individuals is the same as that of historical land use rasters. Figure 1 In this process, connected regions of the same land type among land use planning simulation individuals are denoted as patches, and a maximum number of iterations is set. The fitness value corresponding to each land use planning simulation individual is calculated based on ecological service function, economic output function, spatial structure rationality, and change consistency. Based on the fitness value, the population set is updated and iterated through a genetic algorithm. The update and iteration process includes crossover and recombination and local repair. Local repair is carried out based on a preset rule set. The process continues until the number of update iterations reaches the maximum number of iterations, and the land use planning simulation individual with the highest fitness is output as the land use planning optimization map.

[0008] Preferably, the number of land type changes in the historical land use data sequence is counted to construct a global land transfer matrix. Specifically, this includes the following steps: constructing an N×N empty matrix, where N is the total number of land types; counting the number of land type changes in the historical land use data sequence; filling the i-th row and j-th column of the empty matrix with the ratio of the number of times the i-th land type changes to the j-th land type to the total number of changes, i, j=1, 2, 3, ..., N, until the empty matrix is ​​completely filled; and outputting the empty matrix as the global land transfer matrix.

[0009] Preferably, the land use planning simulation individual is constructed based on the land change probability vector, the land trend probability vector, and the change frequency value, specifically including the following steps:

[0010] Construct a map to be filled, the storage format of which is the same as that of historical land use raster. Figure 1The process involves iterating through the pixels of the map to be filled. For each pixel, it checks whether the frequency of change corresponding to that pixel is higher than a change threshold. If the frequency of change corresponding to that pixel is higher than the threshold, the land type is determined by a roulette wheel selection algorithm based on the land trend probability vector of that pixel and then filled. If the frequency of change corresponding to that pixel is not higher than the threshold, the land type is determined by a roulette wheel selection algorithm based on the land change probability vector of that pixel and then filled. Finally, the filled map is subjected to local repair to obtain a land use planning simulation individual.

[0011] Preferably, the fitness value of each land use planning simulation individual is calculated based on its ecosystem service function, economic output function, spatial structure rationality, and consistency of change. This process specifically includes the following steps:

[0012] The ecological service value and economic output value corresponding to individual land use planning simulations are statistically analyzed.

[0013] Calculate the spatial compactness index, spatial fragmentation index, and neighborhood consistency index corresponding to the simulated land use planning individual, and then perform a weighted summation of the spatial compactness index, spatial fragmentation index, and neighborhood consistency index corresponding to the simulated land use planning individual to obtain the spatial structure rationality score;

[0014] The change consistency is calculated as follows: for each pixel, the probability value of the land type of the individual in the land use planning simulation is determined as the probability vector of the land trend corresponding to the pixel as the pixel change consistency value, and then the average of the change consistency values ​​of all pixels is taken as the change consistency.

[0015] The fitness value is obtained by weighted summing of the ecological service value, economic output value, spatial structure rationality score, and change consistency of the simulated individual in the land use planning.

[0016] Preferably, cross-recombination includes the following steps:

[0017] Randomly select a cell as the target cell, and replace the patch containing the target cell in the two land use planning simulation individuals to be cross-recombined. The specific replacement operation is as follows: designate the two land use planning simulation individuals to be cross-recombined as the first target individual and the second target individual, respectively; designate the patch containing the target cell in the first target individual and the second target individual as the first target patch and the second target patch, respectively; map the second target patch to all cells in the first target individual and reset it; then fill the second target patch into the first target individual; if there are still reset cells in the first target individual, fill them with the land type corresponding to the second target patch; map the first target patch to all cells in the second target individual and reset it; then fill the first target patch into the second target individual; if there are still reset cells in the second target individual, fill them with the land type corresponding to the first target patch.

[0018] Preferably, the preset rule set corresponding to local repair includes the following:

[0019] Iterate through all patches of the land use planning simulation individual. If the area of ​​a patch is smaller than the set patch area of ​​the corresponding land type, select the land type with the largest proportion within 1 pixel width of the patch and replace the patch.

[0020] Iterate through all land types and judge all patches corresponding to each land type. If the shortest path length between two patches is less than the connection threshold, replace the area between the two patches with the land type corresponding to the two patches.

[0021] Iterate through all patches of the land use planning simulation individual. If the patch compactness index of a patch is higher than the set compactness index, replace the land type corresponding to each cell in the patch with the land type with the highest proportion in the 8-neighborhood of that cell.

[0022] This invention also provides a land use planning optimization system based on spatiotemporal big data, comprising:

[0023] The historical land use data sequence acquisition module is used to acquire the historical land use data sequence of the land to be planned. The historical land use data sequence includes a historical land use raster map arranged in chronological order, and each cell in the historical land use raster map stores the land type of the corresponding location of the land to be planned at the corresponding time point in the historical land use raster map.

[0024] The change feature calculation module is used to count the number of land type changes in the historical land use data sequence and construct a global land transfer matrix. For each cell of the land to be planned, the following steps are performed: count the frequency of each land type in the historical land use data sequence, construct the land change probability vector corresponding to the cell, count the number of land type changes in the historical land use data sequence, calculate the change frequency value corresponding to the cell, select the last historical land use raster in the historical land use data sequence, determine the land type of the cell in the selected historical land use raster, and select the row vector in the global land transfer matrix as the land trend probability vector based on the determined land type.

[0025] The iterative optimization module is used to construct a population set composed of several land use planning simulation individuals. These simulation individuals are constructed based on land change probability vectors, land trend probability vectors, and change frequency values. The storage format of the land use planning simulation individuals is the same as that of historical land use rasters. Figure 1 In this process, connected regions of the same land type among land use planning simulation individuals are denoted as patches, and a maximum number of iterations is set. The fitness value corresponding to each land use planning simulation individual is calculated based on ecological service function, economic output function, spatial structure rationality, and change consistency. Based on the fitness value, the population set is updated and iterated through a genetic algorithm. The update and iteration process includes crossover and recombination and local repair. Local repair is carried out based on a preset rule set. The process continues until the number of update iterations reaches the maximum number of iterations, and the land use planning simulation individual with the highest fitness is output as the land use planning optimization map.

[0026] The present invention has the following advantages:

[0027] This invention acquires multi-temporal land use remote sensing raster image sequences of land to be planned, constructs a pixel-level land use change dataset, comprehensively characterizes the land use evolution trends and change sensitivity of each location at different historical stages, and ensures that the model can identify the spatial distribution characteristics of stable areas and high-frequency change areas. Based on land type transformation statistics, a global transition matrix and trend probability vector are constructed to achieve quantitative modeling of land use evolution patterns, which are then used for trend-guided construction of the initial population of the genetic algorithm, improving the directionality and convergence efficiency of optimization. A fitness function is constructed by combining multiple objectives such as ecosystem service value, economic output capacity, spatial structure rationality, and trend consistency to achieve a comprehensive evaluation of ecological protection, economic development, and spatial coordination. During the evolution process, a patch identification mechanism and neighborhood spatial rule control strategy are introduced to maintain the integrity and coherence of the land pattern, and local restoration operations are used to suppress fragmentation and leapfrog development, thereby improving the practical feasibility and spatial rationality of the optimization results. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of the land use planning optimization system based on spatiotemporal big data used in an embodiment of the present invention. Detailed Implementation

[0029] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of this invention.

[0030] Example 1: A land use planning optimization method based on spatiotemporal big data, comprising:

[0031] The historical land use data sequence of the land to be planned is obtained. The historical land use data sequence includes historical land use raster maps arranged in chronological order. Each cell in the historical land use raster map stores the land type of the corresponding location of the land to be planned at the corresponding time point in the historical land use raster map. Land types include forest land, cultivated land, grassland, construction land, and water area, etc., and are generally represented by integer values, such as cultivated land = 1, forest land = 2, grassland = 3, construction land = 4, and water area = 5. It should be noted that the historical land use data sequence generally comes from land use remote sensing interpretation data (such as GlobeLand30 and ESALUD) released by the Ministry of Ecology and Environment of China / Resource Satellite Center. After obtaining the remote sensing image data corresponding to the land to be planned (generally regarded as a rectangle), the corresponding remote sensing image data is rasterized according to the pre-set grid size. Each raster is regarded as a cell and is labeled with the corresponding land type. The time interval between two adjacent historical land use raster maps is generally 5 years.

[0032] The number of land type changes in historical land use data sequences is statistically analyzed to construct a global land transfer matrix. The global land transfer matrix reflects the overall change relationship between different land types and is used to capture the historical evolution pattern of land types, providing trend reference for the initial population construction of subsequent genetic algorithms.

[0033] To count the number of land type changes in historical land use data sequences and construct a global land transfer matrix, the specific steps are as follows: Construct an N×N empty matrix, where N is the total number of land types. Count the number of land type changes in the historical land use data sequences. Fill the i-th row and j-th column of the empty matrix with the ratio of the number of times the i-th land type changes to the j-th land type to the total number of changes, i, j = 1, 2, 3, ..., N, until the empty matrix is ​​completely filled. Output the empty matrix as the global land transfer matrix. It should be noted that the change in land type between adjacent historical land use raster pixels is considered as one land type change. For the sake of data consistency in the global land transfer matrix, for example, the change between cultivated land is also considered as one land type change.

[0034] For each cell of the land to be planned, the following steps are performed: 1. Count the frequency of each land type appearing in the historical land use data sequence, construct the land change probability vector corresponding to the cell, and count the number of land type changes in the cell in the historical land use data sequence, calculating the change frequency value corresponding to the cell. Then, select the last historical land use raster in the historical land use data sequence, determine the land type of the cell in the selected historical land use raster, and select a row vector in the global land transfer matrix as the land trend probability vector based on the determined land type. It should be noted that the number of land type changes here refers to the number of times the land type of corresponding cells in two adjacent historical land use rasters is different, and the change frequency value refers to the ratio of the number of land type changes to the total number of land type changes, which can reflect the degree of change of the land type at the cell's location. 2. Select a row vector in the global land transfer matrix as the land trend probability vector based on the determined land type. Specifically, match the determined land type with the land type corresponding to each row of the global land transfer matrix, and output the row vector corresponding to the row with the successful match.

[0035] A population set is constructed from several land use planning simulation individuals. These individuals are constructed based on land change probability vectors, land trend probability vectors, and change frequency values. The storage format of the land use planning simulation individuals is the same as that of historical land use rasters. Figure 1 In this study, connected regions of the same land type in the land use planning simulation individuals are denoted as patches, and a maximum number of iterations is set. It should be noted that each patch can represent a functional unit of a specific land type. By analyzing the frequency of change and transfer patterns of each cell in historical stages, it is possible to identify which areas are stable in the long term and which areas change frequently, thereby reflecting their spatial sensitivity under natural succession, policy intervention or development pressure. Using this trend information for planning initialization and evolution guidance helps improve the model's fit to the evolution of the real-world pattern, reduces unreasonable changes, and enhances the feasibility and evolutionary consistency of planning schemes. The fitness value of each land use planning simulation individual is calculated based on ecological service functions, economic output functions, spatial structure rationality, and change consistency. Based on the fitness value, a genetic algorithm is used to update and iterate the population set. The update and iteration process includes crossover and recombination, as well as local repair. This update and iteration process is essentially an optimization process for land use planning. Local repair is performed based on a preset rule set. The process continues until the maximum number of iterations is reached. The land use planning simulation individual with the highest fitness is output as the land use planning optimization map, completing the land use planning optimization operation. The land use planning optimization map marks the land types at various locations, providing a reference for land use.

[0036] This invention acquires multi-temporal land use remote sensing raster image sequences of land to be planned, constructs a pixel-level land use change dataset, comprehensively characterizes the land use evolution trends and change sensitivity of each location at different historical stages, and ensures that the model can identify the spatial distribution characteristics of stable areas and high-frequency change areas. Based on land type transformation statistics, a global transition matrix and trend probability vector are constructed to achieve quantitative modeling of land use evolution patterns, which are then used for trend-guided construction of the initial population of the genetic algorithm, improving the directionality and convergence efficiency of optimization. A fitness function is constructed by combining multiple objectives such as ecosystem service value, economic output capacity, spatial structure rationality, and trend consistency to achieve a comprehensive evaluation of ecological protection, economic development, and spatial coordination. During the evolution process, a patch identification mechanism and neighborhood spatial rule control strategy are introduced to maintain the integrity and coherence of the land pattern, and local restoration operations are used to suppress fragmentation and leapfrog development, thereby improving the practical feasibility and spatial rationality of the optimization results.

[0037] The process of constructing a land use planning simulation individual based on the land change probability vector, land trend probability vector, and change frequency value includes the following steps:

[0038] Construct a map to be filled, the storage format of which is the same as that of historical land use raster. Figure 1 The process involves iterating through the pixels of the map to be filled. For each pixel, the frequency of change is checked against a threshold set by the operator. If the frequency of change is higher, the pixel's location has a high frequency of change in historical records. The land type is then determined using a roulette wheel selection algorithm based on the pixel's land trend probability vector for filling. If the frequency of change is lower, the pixel's location has a low frequency of change in historical records. The land type is then determined using a roulette wheel selection algorithm based on the pixel's land change probability vector for filling. Finally, the filled map is partially repaired to obtain a simulated land use plan individual.

[0039] It should be noted that for pixels with frequent historical land type changes, sampling using historical distribution histograms (i.e., filling in the pixels with land type by determining the land type based on the corresponding land trend probability vector using a roulette wheel selection algorithm) can reflect the diversity and spatial uncertainty of their evolution, enhance the diversity and generalization ability of the initial solutions, and provide a wider search starting point for genetic algorithms. For pixels with low change frequency, sampling using land change probability vectors ensures that their initial land type configuration is closer to the evolution trend, avoids unreasonable disturbances to stable areas, and improves the feasibility and evolutionary direction of the initial individuals from the source.

[0040] The fitness value for each land use planning simulation individual is calculated based on ecological service function, economic output function, spatial structure rationality, and consistency of change. The specific steps include the following:

[0041] This section describes the ecosystem service value and economic output value corresponding to the simulated individual land use plan. It should be noted that during the implementation of land use planning optimization, operators generally set corresponding ecosystem service values ​​and economic output values ​​for each land type, for example, by referring to the "Ecosystem Service Value Equivalent Table" (Xie Gaodi et al.) and the "Statistical Yearbook" published by the National Bureau of Statistics. This item represents the ecosystem service function and economic output function corresponding to the simulated individual land use plan.

[0042] The spatial compactness index, spatial fragmentation index, and neighborhood consistency index corresponding to the land use planning simulation individual are calculated. Then, the spatial compactness index, spatial fragmentation index, and neighborhood consistency index corresponding to the land use planning simulation individual are weighted and summed to obtain the spatial structure rationality score. The spatial structure rationality score is used to characterize the rationality of the spatial structure. It should be noted that the weights in the process of performing the weighted summation are set by experts or by methods such as AHP.

[0043] The spatial compactness index is calculated as follows: For each patch, the ratio of the square of the patch's perimeter to the patch's area is calculated and recorded as the patch compactness index. This index is used to measure whether the patch forms a compact cluster, whether the boundary is regular, and whether there is any "jump" expansion. The average of the patch compactness indices of all patches is then recorded as the spatial compactness index.

[0044] The spatial fragmentation index is calculated as follows: For each land type, the inverse of the ratio of the number of patches corresponding to the land type to the total area corresponding to the land type is recorded as the land type fragmentation index, which is used to measure whether a certain land type is excessively divided into many small fragments, affecting its functional integrity. The average of the land type fragmentation indices of all land types is recorded as the spatial fragmentation index.

[0045] The neighborhood consistency index is calculated as follows: for each cell, the ratio of the number of cells in its 8-neighborhood that have the same land type as the center cell to the total number of cells in its 8-neighborhood is selected as the neighborhood consistency index, which is used to measure the degree of land type clustering in a local area.

[0046] The change consistency is calculated as follows: for each pixel, the probability value of the land type of the individual in the land use planning simulation is determined as the probability vector of the land trend corresponding to the pixel as the pixel change consistency value, and then the average of the change consistency values ​​of all pixels is taken as the change consistency.

[0047] The fitness value is obtained by weighted summing of the ecological service value, economic output value, spatial structure rationality score, and change consistency of the simulated individual in the land use planning.

[0048] Cross-recombination includes the following steps:

[0049] Randomly select a cell as the target cell, and replace the patch containing the target cell in the two land use planning simulation individuals to be cross-recombined. The specific replacement operation is as follows: designate the two land use planning simulation individuals to be cross-recombined as the first target individual and the second target individual, respectively; designate the patch containing the target cell in the first target individual and the second target individual as the first target patch and the second target patch, respectively; map the second target patch to all cells in the first target individual and reset it; then fill the second target patch into the first target individual; if there are still reset cells in the first target individual, fill them with the land type corresponding to the second target patch; map the first target patch to all cells in the second target individual and reset it; then fill the first target patch into the second target individual; if there are still reset cells in the second target individual, fill them with the land type corresponding to the first target patch.

[0050] The preset rule set for local repair includes the following:

[0051] Iterate through all patches of the land use planning simulation individual. If the area of ​​a patch is smaller than the set patch area for the corresponding land type, select the land type with the largest proportion within 1 pixel width of the patch and replace the patch. It should be noted that the operator will set a set patch area for each land type based on experience.

[0052] Iterate through all land types and judge all patches corresponding to each land type. If the shortest path length between two patches is less than the connection threshold (set by the operator), replace the area between the two patches with the land type corresponding to the two patches.

[0053] Iterate through all patches of the land use planning simulation individual. If the patch compactness index of a patch is higher than the set compactness index (the set compactness index is determined by the operator), replace the land type corresponding to each cell in the patch with the land type with the highest proportion in the 8-neighborhood of that cell.

[0054] Example 2: A land use planning optimization system based on spatiotemporal big data, see [link / reference] Figure 1 ,include:

[0055] The historical land use data sequence acquisition module is used to acquire the historical land use data sequence of the land to be planned. The historical land use data sequence includes historical land use raster maps arranged in chronological order. Each raster in the historical land use raster map stores the land type of the corresponding location of the land to be planned at the corresponding time point in the historical land use raster map. Land types include forest land, cultivated land, grassland, construction land, and water area, etc., and are generally represented by integer values, for example, cultivated land = 1, forest land = 2, grassland = 3, construction land = 4, and water area = 5. It should be noted that the historical land use data sequence generally comes from land use remote sensing interpretation data (such as GlobeLand30 and ESALUD) released by the Ministry of Ecology and Environment of China / Resource Satellite Center. After acquiring the remote sensing image data corresponding to the land to be planned (generally regarded as a rectangle), the corresponding remote sensing image data is rasterized according to the pre-set grid size. Each raster is regarded as a pixel area and is marked with the corresponding land type. The time interval between two adjacent historical land use raster maps is generally 5 years.

[0056] The change feature calculation module performs the following steps for each cell of the land to be planned: It counts the frequency of each land type appearing in the historical land use data sequence, constructs the land change probability vector corresponding to the cell, counts the number of land type changes in the cell in the historical land use data sequence, calculates the change frequency value corresponding to the cell, selects the last historical land use raster in the historical land use data sequence, determines the land type of the cell in the selected historical land use raster, and selects a row vector in the global land transfer matrix as the land trend probability vector based on the determined land type. It should be noted that the number of land type changes here refers to the number of times the land type of corresponding cells in two adjacent historical land use rasters is different, and the change frequency value refers to the ratio of the number of land type changes to the total number of land type changes, which can reflect the degree of change of the land type at the cell's location. Specifically, it matches the determined land type with the land type corresponding to each row of the global land transfer matrix and outputs the row vector corresponding to the row with the successful match.

[0057] The iterative optimization module is used to construct a population set composed of several land use planning simulation individuals. These simulation individuals are constructed based on land change probability vectors, land trend probability vectors, and change frequency values. The storage format of the land use planning simulation individuals is the same as that of historical land use rasters. Figure 1In this study, connected regions of the same land type in the land use planning simulation individuals are denoted as patches, and a maximum number of iterations is set. It should be noted that each patch can represent a functional unit of a specific land type. By analyzing the frequency of change and transfer patterns of each cell in historical stages, it is possible to identify which areas are stable in the long term and which areas change frequently, thereby reflecting their spatial sensitivity under natural succession, policy intervention or development pressure. Using this trend information for planning initialization and evolution guidance helps improve the model's fit to the evolution of the real-world pattern, reduces unreasonable changes, and enhances the feasibility and evolutionary consistency of planning schemes. The fitness value of each land use planning simulation individual is calculated based on ecological service functions, economic output functions, spatial structure rationality, and change consistency. Based on the fitness value, a genetic algorithm is used to update and iterate the population set. The update and iteration process includes crossover and recombination, as well as local repair. This update and iteration process is essentially an optimization process for land use planning. Local repair is performed based on a preset rule set. The process continues until the maximum number of iterations is reached. The land use planning simulation individual with the highest fitness is output as the land use planning optimization map, completing the land use planning optimization operation. The land use planning optimization map marks the land types at each location, providing a reference for land use.

[0058] It should be understood that those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims. Parts not described in detail in this specification are prior art known to those skilled in the art.

Claims

1. A land use planning optimization method based on spatiotemporal big data, characterized in that, include: Obtain the historical land use data sequence of the land to be planned. The historical land use data sequence includes historical land use raster maps arranged in chronological order, and each cell in the historical land use raster map stores the land type of the corresponding location of the land to be planned at the corresponding time point in the historical land use raster map. The number of land type changes in the historical land use data sequence is counted to construct a global land transfer matrix. For each cell of the land to be planned, the following steps are performed: the frequency of each land type appearing in the cell in the historical land use data sequence is counted, the land change probability vector corresponding to the cell is constructed, the number of land type changes in the cell in the historical land use data sequence is counted, the change frequency value corresponding to the cell is calculated, the last historical land use raster in the historical land use data sequence is selected, the land type of the cell in the selected historical land use raster is determined, and the row vector in the global land transfer matrix is ​​selected as the land trend probability vector based on the determined land type. A population set is constructed by constructing several land use planning simulation individuals. The land use planning simulation individuals are constructed based on the land change probability vector, the land trend probability vector, and the change frequency value. The storage format of the land use planning simulation individuals is consistent with the historical land use raster map. Connected regions with the same land type in the land use planning simulation individuals are recorded as patches, and the maximum number of iterations is set. The fitness value of each land use planning simulation individual is calculated based on ecological service function, economic output function, spatial structure rationality and change consistency. Based on the fitness value, the population set is updated iteratively using a genetic algorithm. The update iterative process includes crossover and recombination and local repair. Local repair is performed based on a preset rule set. The update iterative process continues until the maximum number of iterations is reached. The land use planning simulation individual with the highest fitness is then output as the land use planning optimization map. The process of constructing a land use planning simulation individual based on the land change probability vector, land trend probability vector, and change frequency value includes the following steps: A map to be filled is constructed, and its storage format is consistent with the historical land use raster map. The cells of the map to be filled are traversed. For each traversed cell, it is determined whether the change frequency value corresponding to the cell is higher than the change threshold. If the change frequency value corresponding to the cell is higher than the change threshold, the land type is determined by the roulette wheel selection algorithm based on the land trend probability vector corresponding to the cell and then filled. If the change frequency value corresponding to the cell is not higher than the change threshold, the land type is determined by the roulette wheel selection algorithm based on the land change probability vector corresponding to the cell and then filled. Then, the filled map to be filled is subjected to local repair to obtain the land use planning simulation individual. The fitness value for each land use planning simulation individual is calculated based on ecological service function, economic output function, spatial structure rationality, and consistency of change. The specific steps include the following: The ecological service value and economic output value corresponding to individual land use planning simulations are statistically analyzed. Calculate the spatial compactness index, spatial fragmentation index, and neighborhood consistency index corresponding to the simulated land use planning individual, and then perform a weighted summation of the spatial compactness index, spatial fragmentation index, and neighborhood consistency index corresponding to the simulated land use planning individual to obtain the spatial structure rationality score; The change consistency is calculated as follows: for each pixel, the probability value of the land type of the individual in the land use planning simulation is determined as the probability vector of the land trend corresponding to the pixel as the pixel change consistency value, and then the average of the change consistency values ​​of all pixels is taken as the change consistency. The fitness value is obtained by weighted summing of the ecological service value, economic output value, spatial structure rationality score, and change consistency of the simulated individual in the land use planning.

2. The land use planning optimization method based on spatiotemporal big data according to claim 1, characterized in that, To count the number of land type changes in historical land use data sequences and construct a global land transfer matrix, the specific steps are as follows: Construct an N×N empty matrix, where N is the total number of land types; count the number of land type changes in historical land use data sequences; fill the i-th row and j-th column of the empty matrix with the ratio of the number of times the i-th land type changes to the j-th land type to the total number of changes, i, j=1, 2, 3, ..., N, until the empty matrix is ​​completely filled; output the empty matrix as the global land transfer matrix.

3. The land use planning optimization method based on spatiotemporal big data according to claim 2, characterized in that, Cross-recombination includes the following steps: Randomly select a cell as the target cell, and replace the patch containing the target cell in the two land use planning simulation individuals to be cross-recombined. The specific replacement operation is as follows: designate the two land use planning simulation individuals to be cross-recombined as the first target individual and the second target individual, respectively; designate the patch containing the target cell in the first target individual and the second target individual as the first target patch and the second target patch, respectively; map the second target patch to all cells in the first target individual and reset it; then fill the second target patch into the first target individual; if there are still reset cells in the first target individual, fill them with the land type corresponding to the second target patch; map the first target patch to all cells in the second target individual and reset it; then fill the first target patch into the second target individual; if there are still reset cells in the second target individual, fill them with the land type corresponding to the first target patch.

4. The land use planning optimization method based on spatiotemporal big data according to claim 3, characterized in that, The preset rule set for local repair includes the following: Iterate through all patches of the land use planning simulation individual. If the area of ​​a patch is smaller than the set patch area of ​​the corresponding land type, select the land type with the largest proportion within 1 pixel width of the patch and replace the patch. Iterate through all land types and judge all patches corresponding to each land type. If the shortest path length between two patches is less than the connection threshold, replace the area between the two patches with the land type corresponding to the two patches. Iterate through all patches of the land use planning simulation individual. If the patch compactness index of a patch is higher than the set compactness index, replace the land type corresponding to each cell in the patch with the land type with the highest proportion in the 8-neighborhood of that cell.

5. A land use planning optimization system based on spatiotemporal big data, characterized in that, The system applies a land use planning optimization method based on spatiotemporal big data as described in any one of claims 1-4, including: The historical land use data sequence acquisition module is used to acquire the historical land use data sequence of the land to be planned. The historical land use data sequence includes a historical land use raster map arranged in chronological order, and each cell in the historical land use raster map stores the land type of the corresponding location of the land to be planned at the corresponding time point in the historical land use raster map. The change feature calculation module is used to count the number of land type changes in the historical land use data sequence and construct a global land transfer matrix. For each cell of the land to be planned, the following steps are performed: count the frequency of each land type in the historical land use data sequence, construct the land change probability vector corresponding to the cell, count the number of land type changes in the historical land use data sequence, calculate the change frequency value corresponding to the cell, select the last historical land use raster in the historical land use data sequence, determine the land type of the cell in the selected historical land use raster, and select the row vector in the global land transfer matrix as the land trend probability vector based on the determined land type. The iterative optimization module constructs a population set composed of several land use planning simulation individuals. These simulation individuals are constructed based on land change probability vectors, land trend probability vectors, and change frequency values. The storage format of the simulation individuals is consistent with the historical land use raster map. Connected regions of the same land type within the simulation individuals are designated as patches, and a maximum number of iterations is set. The module calculates the fitness value for each simulation individual based on ecosystem service function, economic output function, spatial structure rationality, and change consistency. Based on the fitness values, the population set is updated iteratively using a genetic algorithm. The update and iteration process includes crossover and recombination, and local repair. Local repair is performed based on a preset rule set. The process continues until the maximum number of iterations is reached, at which point the land use planning simulation individual with the highest fitness is output as the land use planning optimization map.

Images