Land space planning intelligent optimization method and system based on multi-source data fusion

Abstract

The invention discloses an intelligent optimization method and system for territorial space planning based on multi-source data fusion, and belongs to the technical field of territorial space planning. Constructing a data consistency evaluation model, and generating a confidence weight matrix; local sensitivity analysis is carried out, and space conflict units are identified; generating a minimum disturbance path and a feasibility sequence based on the neighborhood stability and the historical change frequency; searching an optimal path combination through a multi-objective optimization algorithm; and identifying a high-sensitivity area as a key planning weak-sensitivity area, executing a dynamic fine tuning strategy, and finally outputting an optimized territorial space layout result. The method can effectively improve the scientificity and adaptability of space planning, and has good practical application value.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of land space planning, in particular to a land space planning intelligent optimization method and system based on multi-source data fusion. BACKGROUND

[0002] With the advancement of new urbanization and ecological civilization construction, land space planning is becoming increasingly complex, requiring more precision and intelligence. However, most current land space planning still mainly relies on single source data (such as remote sensing images or population statistics) for land suitability evaluation and spatial layout optimization, making it difficult to effectively integrate multi-source heterogeneous data such as transportation networks, ecological sensitive areas, social and economic indicators, and historical planning conflict areas, resulting in problems such as local incoordination of spatial planning schemes, low resource allocation efficiency, and repeated correction of ecological sensitive conflict points.

[0003] Especially in the process of urban expansion in hilly and mountainous areas, due to complex constraints such as terrain slope, ecological red line, and historical construction overlap, traditional methods are difficult to balance between global coordination and local fine-tuning, resulting in repeated overturning and recalculation of some planning units, wasting a large amount of human resources and time cost. For example, in a city renewal project in a mountainous city edge area, four types of data constraints such as landslide hazard area, farmland protection area, historical and cultural block, and highway corridor need to be considered. Due to the differences in resolution and update frequency of various data, existing methods cannot accurately construct a unified spatial decision model, resulting in multiple layout conflicts and planning lag, which seriously restricts urban development. SUMMARY

[0004] The purpose of the present application is to provide a land space planning intelligent optimization method and system based on multi-source data fusion to solve the problems in the background art.

[0005] In order to achieve the above-mentioned purpose, the present application provides the following technical solution: a land space planning intelligent optimization method based on multi-source data fusion, comprising:

[0006] obtaining a multi-source spatial data set D in the target planning area;

[0007] based on the spatial resolution, timeliness and coordinate consistency parameters in the data set D, constructing a multi-source data consistency evaluation model R, and outputting a multi-source data fusion confidence weight matrix M;

[0008] According to the weight matrix M, local sensitivity analysis is performed on the target area, and a set of spatial conflict units is identified;

[0009] For each conflict unit, based on its neighborhood pattern stability and historical change frequency, a minimum disturbance space transformation path L(c) is constructed, and a feasible transformation sequence T(c) is generated;

[0010] Based on the planning objective function and constraint conditions, intelligent search optimization is performed on the potential scheme space S composed of all T(c) combinations to obtain an optimal conversion path set T*, wherein the planning objective function includes the maximization of ecological protection integrity, the maximization of land use efficiency and the minimization of planning modification cost;

[0011] The characteristic disturbance region in the optimal path set T* is labeled as a key planning weak sensitive region, and an adjustment strategy is performed on the Z region, including dynamic weight adjustment, neighborhood reclassification and elastic land reservation, and an optimized national space layout result is output.

[0012] Preferably, the multi-source spatial data set D includes terrain slope data, land use status data, ecological sensitive area vector boundary, traffic network structure, historical planning boundary and construction land approval record.

[0013] Preferably, the multi-source data consistency evaluation model R is constructed, including:

[0014] For each data layer in the multi-source data set D, the original spatial resolution parameter δ_i, the data acquisition timestamp τ_i and the projection coordinate system information π_i are extracted to form a data original meta-information set {δ_i, τ_i, π_i}, wherein i is the layer number;

[0015] Based on the proportional relationship between the spatial resolution parameter δ_i and the target unified resolution δ0, the spatial precision deviation factor α_i is calculated;

[0016] According to the time difference Δτ_i between the data acquisition timestamp τ_i and the current time τ_now, an exponential decay function is used to construct a timeliness weight factor β_i, wherein k is an empirical coefficient;

[0017] Based on the consistency relationship between the projection coordinate system π_i and the unified target coordinate system π0, a hash comparison method is used to judge whether they are completely consistent, if they are consistent, γ_i = 1 is assigned, if they need to be re-projected and converted, γ_i ∈ (0, 1) is calculated according to the boundary offset rate of the converted layer and the original layer to form a coordinate consistency weight factor;

[0018] According to the factors α_i, β_i and γ_i, a consistency comprehensive evaluation model R_i is constructed, and its expression is: ; wherein λ1, λ2 and λ3 are normalized weight coefficients, satisfying ;

[0019] After the R_i values of all layers are normalized, a confidence weight matrix M = {w1, w2,..., w_n} is formed, wherein .

[0020] Preferably, the identifying the set of spatial conflict units comprises:

[0021] Stacking each data layer in the weight matrix M according to the grid unit, calculating the comprehensive sensitivity index W(x, y) of each unit;

[0022] Setting a spatial conflict identification threshold θ, when W(x, y) exceeds θ and there are more than two high-priority conflict attributes, marking the unit as a candidate conflict unit;

[0023] Based on the eight-neighborhood analysis method, the connectivity clustering of the candidate conflict units is carried out, and the scattered units with an area less than a set area threshold are removed;

[0024] Output the set of spatial conflict units C after screening.

[0025] Preferably, the constructing the minimum disturbance spatial transformation path L(c) and generating the feasible transformation sequence T(c) comprises:

[0026] Based on the spatial position of the conflict unit c, extracting the spatial structure type within its 3-order neighborhood, calculating the neighborhood pattern stability S(c), defined as the proportion of the dominant land type in the neighborhood;

[0027] Querying the historical land use records of the conflict unit c and its neighborhood, counting the number of land type changes in the past N years, calculating the historical change frequency H(c), and constructing the disturbance sensitivity score R(c) based on S(c) and H(c);

[0028] According to R(c), the conflict units are divided into disturbance acceptable levels, and the transformation path set L(c) under the minimum disturbance principle is limited, which is the path set from the current land use state to other planning allowed states, and the path step is controlled by the planning intensity and ecological limit;

[0029] Combined with the current policy guidance and regional planning target, the path feasibility screening and sorting of L(c) are carried out, and the feasible transformation sequence T(c) is generated, wherein the paths in T(c) are sorted according to the principle of minimum comprehensive cost.

[0030] Preferably, the obtaining the optimal transformation path set T* comprises:

[0031] Constructing a potential scheme space S, performing Cartesian combination of the feasible transformation sequence T(c) of each conflict unit c according to the serial number, forming a potential scheme space S, and recording each scheme as sk, which contains a transformation decision vector for the whole set C;

[0032] Define the mathematical representation of the planning objective function F and the constraints G. F is a set of multi-objective functions, including maximizing the integrity of ecological protection, maximizing land use efficiency, and minimizing the planning correction cost. It is represented by a weighted linear or Pareto dominance scheme. G includes hard constraints and soft constraints, and is embedded in the optimization process as a feasibility criterion.

[0033] A hybrid intelligent search algorithm is used to optimize S: a hybrid strategy combining multi-objective evolutionary algorithm and local search is preferred. In each generation of computation, the candidate schemes are constrained, the penalty function is evaluated, and the multi-objective ranking is performed. The applicable encoding is binary or integer vector. The termination criteria are set as the maximum number of generations, the objective convergence threshold, or the number of iterations without improvement.

[0034] Output the optimal transformation path set T*: Select several paths from the final Pareto front or weighted optimal solution according to the decision-maker's preference to form T*, and calculate the detailed cost, suitability and implementation priority for each selected path.

[0035] Preferably, the step of marking the feature perturbation region in the optimal path set T* as a critical planning weakly sensitive region and implementing an adjustment strategy for region Z includes:

[0036] Based on the disturbance sensitivity score R(c) of each conflict unit in the optimal path set T*, regions with sensitivity scores higher than the set threshold σ are selected, and continuous high-sensitivity units are aggregated into feature disturbance regions by spatial clustering algorithm, which are marked as key planning weakly sensitive regions Z.

[0037] For each weakly sensitive region Z, calculate the neighborhood structure continuity index and resource carrying capacity index, and construct the multidimensional feature matrix F(Z) of the weakly sensitive region.

[0038] Based on the F(Z) results and regional planning objectives, a dynamic weight adjustment strategy is adopted to adjust the land use function of area Z, including reallocating land use grades, adjusting land use ratios, or reserving flexible land use boundaries.

[0039] Iterative verification is performed on the adjusted Z region. The impact of the adjustment scheme on the global planning objective function F is evaluated using simulated annealing or multi-objective local search algorithms. The final adjustment scheme is then output to the planning result layer.

[0040] This invention also provides an intelligent optimization system for land spatial planning based on multi-source data fusion, comprising:

[0041] Multi-source spatial data acquisition module: Acquires multi-source spatial data set D within the target planning area;

[0042] Weighted modeling module: Based on the spatial resolution, timeliness and coordinate consistency parameters in the dataset D, construct a multi-source data consistency evaluation model R, and output the multi-source data fusion confidence weight matrix M;

[0043] Conflict identification module: Based on the weight matrix M, it performs local sensitivity analysis on the target area and identifies the set of spatial conflict units C;

[0044] Path generation module: For each conflict unit, based on the stability of its neighborhood pattern and the frequency of historical changes, construct the minimum disturbance spatial transformation path L(c) and generate a feasible transformation sequence T(c);

[0045] Optimal path set construction module: Based on the planning objective function and constraints, intelligent search and optimization are performed on the potential solution space S composed of all T(c) combinations to obtain the optimal transformation path set T*, where the planning objective function includes the objectives of maximizing ecological protection integrity, maximizing land use efficiency and minimizing planning correction costs;

[0046] Dynamic adjustment module: Marks the feature disturbance areas in the optimal path set T* as key planning weakly sensitive areas, and performs adjustment strategies on area Z, including dynamic weight adjustment, neighborhood reclassification and flexible land use reservation, and outputs the optimized land spatial layout results.

[0047] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0048] 1. This invention systematically solves the problems of difficulty in integrating heterogeneous data and strong subjectivity in judging data quality in existing land spatial planning by constructing a consistency evaluation model and confidence weight matrix based on multi-source spatial data. It achieves quantitative weight control of multi-source data in spatial, temporal, and coordinate dimensions, thus providing a stable and controllable data foundation for subsequent planning conflict identification and optimization. Compared with traditional methods relying on single or manual layer selection, this invention significantly improves the objectivity and automation level of data processing.

[0049] 2. This invention is the first to propose constructing a minimum disturbance transformation path based on disturbance sensitivity scoring, and introduces an intelligent search optimization algorithm to combine and solve multi-objective solutions for spatial conflicts, ultimately achieving a dynamic trade-off between ensuring ecological protection, improving land efficiency, and reducing the cost of planning corrections. By introducing a key planning weak-sensitivity area identification and fine-tuning mechanism, flexible regulation of highly sensitive areas is achieved, effectively improving the scientific nature, precision, and feasibility of national land spatial layout. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0051] Figure 1 This is a flowchart of the method of the present invention.

[0052] Figure 2 This is a flowchart of the system modules of the present invention. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] Example 1, please refer to Figure 1 As shown in this embodiment, the intelligent optimization method for land spatial planning based on multi-source data fusion includes:

[0055] Obtain a multi-source spatial data set D within the target planning area, including topographic slope data, current land use data, vector boundaries of ecologically sensitive areas, transportation network structure, historical planning boundaries, and construction land approval records;

[0056] Based on the spatial resolution, timeliness, and coordinate consistency parameters in the dataset D, a multi-source data consistency evaluation model R is constructed, which outputs a multi-source data fusion confidence weight matrix M to measure the credibility and participation of various types of data in spatial decision-making.

[0057] Based on the weight matrix M, a local sensitivity analysis is performed on the target area to identify a set of spatial conflict units C. The spatial conflict unit C is a grid unit that falls into at least two types of high-priority land use or ecological red lines at the same time.

[0058] For each conflict unit, based on the stability of its neighborhood pattern and the frequency of historical changes, a minimum disturbance spatial transformation path L(c) is constructed, and a feasible transformation sequence T(c) is generated.

[0059] Based on the planning objective function and constraints, intelligent search and optimization are performed on the potential solution space S composed of all T(c) combinations to obtain the optimal transformation path set T*, where the planning objective function includes the objectives of maximizing ecological protection integrity, maximizing land use efficiency and minimizing planning correction costs.

[0060] The characteristic perturbation areas in the optimal path set T* are marked as key planning weakly sensitive areas, and a fine-tuning strategy is implemented for area Z, including dynamic weight adjustment, neighborhood reclassification and flexible land use reservation, and the optimized land spatial layout results are output.

[0061] This invention provides an intelligent optimization method for land spatial planning based on multi-source data fusion. First, it requires acquiring a multi-source spatial data set D within the target planning area, which serves as the data foundation for the entire optimization process. The multi-source spatial data set D includes, but is not limited to, the following six main types of spatial information data:

[0062] Topographic slope data reflects the degree of surface undulation in a region and is mainly derived from a high-resolution digital elevation model (DEM). Typically, remote sensing mapping data with a resolution better than 30 meters is selected for acquisition. The slope value is obtained by calculating the elevation change rate of each raster cell in the DEM and then performing numerical normalization from 0° to 90° for subsequent terrain suitability analysis.

[0063] Land use status data: This describes the actual land use type within the current area, such as residential land, industrial land, agricultural land, water areas, and forest land. The data typically originates from local natural resources departments or the National Land Survey Database, and is in raster or vector format. It needs to be uniformly resampled to the target analysis resolution (e.g., 30 meters or 100 meters) before processing.

[0064] Vector boundaries of ecologically sensitive areas: These include basic ecological control lines, drinking water source protection areas, nature reserves, wetland red lines, and geologically hazardous areas. This data is mostly in vector format and requires spatial rasterization. Sensitive area units are marked as unplannable or have a high constraint level.

[0065] Transportation network structure data includes the spatial paths of transportation infrastructure networks such as highways, railways, urban arterial roads, and rail transit. This data is used to analyze accessibility indicators and development axis structures, and is usually expressed in the form of vector line data. The traffic influence index of each unit can be calculated through spatial neighborhood.

[0066] Historical planning boundary data refers to land planning data that has been approved and implemented in the past but has not yet been completed or abandoned. It includes historical master plan control boundaries, development intensity indicators, land use codes, etc. The purpose is to assess the coupling and conflict between the existing planning basis and the new plan.

[0067] Construction land approval records: Obtain a list of construction land project approvals from the past 5 to 10 years, including location coordinates, project type, approval time, actual development progress, etc., to help determine the authenticity and evolution trend of land demand.

[0068] To achieve unified processing of data fusion, the following standardized procedures need to be performed on the above six types of data:

[0069] All data must be unified to the same projection coordinate system, preferably the Chinese Geodetic Coordinate System 2000 (CGCS2000). Raster and vector data should be uniformly reprojected, and spatial alignment should be performed using bilinear interpolation or nearest neighbor interpolation to ensure consistent spatial coverage boundaries, with errors controlled within one raster unit.

[0070] To enable collaborative application of multi-source data within the same spatial model, all input data needs to be converted into a raster format with a uniform spatial resolution (e.g., 30m x 30m), and normalization processing needs to be performed on various attribute values. Taking slope as an example, the original slope data ranges from 0 to 90 degrees, and normalization is achieved by dividing by the maximum value of 90. Land use data is processed through assignment, such as assigning a value of 0.8 to residential land, 0.3 to ecological land, and 0.6 to industrial land.

[0071] The final dataset D is a fused dataset containing multiple spatial layers, each of which has been normalized, rasterized, and assigned confidence weights, serving as the basic input for subsequent planning conflict identification, sensitive area location, and intelligent optimization.

[0072] To achieve effective fusion of multi-source heterogeneous spatial data, after obtaining the data set D, this invention proposes a method for constructing a multi-source data consistency evaluation model R, which is used to quantify the credibility and participation weight of various types of data in the fusion process, and generates a multi-source data fusion confidence weight matrix M accordingly, which serves as the input basis for subsequent planning conflict identification and spatial optimization calculation.

[0073] The input is a dataset D = {D1, D2, ..., Dn}, where each Di represents the i-th type of spatial data layer and contains the following metadata:

[0074] Spatial resolution δ_i (unit: meters)

[0075] Collect timestamp τ_i (format: YYYY-MM-DD)

[0076] Coordinate system information π_i (e.g., CGCS2000, WGS84, etc.)

[0077] The goal is to construct a set of weight values ​​M={w1, w2, ..., wn}, where wi is the confidence weight of the i-th data layer in the fusion process.

[0078] The consistency model R is modeled from three dimensions: spatial resolution adaptability (α), timeliness (β), and coordinate system consistency (γ). Each dimension is expressed in the form of a quantization function, which ultimately constitutes the comprehensive evaluation model R_i.

[0079] Spatial resolution is a core indicator for measuring data accuracy. Assuming the standard fusion resolution used in this invention's system is δ0 (e.g., set to 30 meters), the spatial accuracy deviation factor α_i for the i-th layer is calculated as follows:

[0080] When δ_i ≥ δ0, set ;

[0081] When δ_i < δ0, set α_i = 1;

[0082] This design aims to limit the weight inflation of high-precision data during the fusion process, preventing it from misleading the planning model.

[0083] Because the data collection times of each data layer differ significantly, it is necessary to measure the "freshness" of the data. Let the current system running date be τ_now, and the collection time of the i-th type of layer be τ_i, then the time difference is Δτ_i = τ_now - τ_i, measured in "years".

[0084] The time decay trend is represented by an exponential decay function, defined as follows: β_i = exp(-k × Δτ_i); where k is the time decay factor, and a recommended value range is 0.2 to 0.5. Taking k = 0.3 as an example, when the data was collected 2 years ago, the value of β_i is approximately 0.55; if it was collected 5 years ago, then β_i drops to approximately 0.22.

[0085] If the coordinate system π_i used by the layer is completely consistent with the target unified coordinate system π0, then γ_i = 1; otherwise, a projection transformation is required. After the transformation, the change in the boundary envelope is compared. Let the boundary offset rate be δ_proj_i, then we can get: γ_i = max(0, 1 - δ_proj_i / ε); where ε is the coordinate offset threshold (the default setting is 30 meters). When δ_proj_i exceeds the threshold, γ_i drops to 0; if there is only a small error, γ_i is close to 1.

[0086] After obtaining the three factors α_i, β_i, and γ_i, a weighted summation model is used to construct the consistency score R_i, as shown in the formula: Where λ1, λ2, and λ3 represent the weight coefficients of the three dimensions, satisfying... To ensure balanced evaluation, it is recommended to set λ1 = 0.4, λ2 = 0.3, and λ3 = 0.3, meaning that spatial resolution adaptability takes precedence over other dimensions. The value of R_i ranges from 0 to 1; a higher value indicates that the data in that layer better meets the integration requirements of the current spatial planning model.

[0087] After completing the R_i scoring for all layers, to ensure weight normalization, a linear normalization method is used to generate the confidence weight matrix M, calculated as follows: , where j = 1 to n; finally, we get M = {w1,w2, ..., wn}, which sums to 1, and each w_i represents the degree of fusion participation of the i-th layer in the subsequent model.

[0088] The core of local sensitivity analysis lies in quantifying the constraint strength and planning risk borne by each spatial unit (i.e., raster) under different data layers. By weighting and fusing the raster values ​​of multi-source data with their corresponding confidence weights, a comprehensive sensitivity index W(x, y) can be obtained, which represents the sensitivity of the raster unit to the uncertainty or conflict of spatial planning schemes.

[0089] Suppose the spatial analysis region is divided into N×N raster cells, and the coordinates of any cell are denoted as (x, y). M = {w1, w2, ..., wn} is the confidence weight matrix of the multi-source layers, and V_i(x, y) is the normalized raster value of the i-th layer at position (x, y). Then, the comprehensive sensitivity index W(x, y) is calculated as follows: W(x, y) = ∑ (i = 1 to n) [w_i × V_i(x, y)]; that is, the sensitivity score of each cell is obtained by weighted summation of the raster values ​​of all layers. W(x, y) ∈ [0, 1], and a higher value indicates greater instability or a higher degree of constraint.

[0090] This invention combines highly sensitive gratings with attribute conflict overlay analysis, and proposes the following conflict cell identification steps:

[0091] To identify high-risk areas, a sensitivity threshold θ is set, and it is recommended to set it to the 85th percentile of W_max, that is, retaining high-value units with sensitivity scores in the top 15%. Units that satisfy W(x, y)≥θ are considered candidate conflict units.

[0092] Furthermore, units with overlapping conflict attributes are further screened. If V_i(x, y) and V_j(x, y) at position (x, y) are both high priority (e.g., ecological red lines and transportation corridors overlap), then the unit is considered a unit with overlapping conflict attributes. Units that only meet the criteria of high sensitivity but do not have conflict attributes will be eliminated.

[0093] Spatial connectivity clustering is performed on all candidate conflict units, using an eight-neighbor clustering method (i.e., up, down, left, right, and four diagonal directions) to group spatially adjacent conflict units into the same conflict patch. To prevent misidentification or noise interference, a minimum patch area threshold A_min is set, typically three consecutive grid cells (areas smaller than 2700 square meters are considered invalid).

[0094] The set of spatial units after screening and clustering is defined as the set of spatial conflict units C, denoted as: C = {c1,c2, ..., ck}; where each cj represents a spatial conflict region with high sensitivity and attribute conflict characteristics.

[0095] After identifying the set of spatial conflict units C, this invention further proposes a method for constructing the minimum disturbance spatial transformation path L(c) for each conflict unit c. By analyzing the stability of its neighborhood pattern S(c) and the frequency of historical changes H(c), the disturbance sensitivity score R(c) is calculated to achieve a reasonable transformation of land use with the minimum cost, and finally outputs a feasible transformation sequence T(c) to provide a set of target schemes for subsequent optimization of national land spatial layout.

[0096] Traditional land use planning often involves direct land use adjustments to conflicting units, neglecting the spatial stability and historical evolution trends of their respective areas. This frequently leads to problems such as fragmented local land use, disrupted ecological patterns, or plans that cannot be implemented. To address these challenges, this invention introduces "neighborhood pattern stability" and "frequency of historical changes" as criteria for judging disturbance tolerance. It prioritizes path design within areas where disturbance is acceptable, avoiding strong intervention in areas with stable patterns or sensitive to changes, thus achieving minimal spatial disturbance adjustments.

[0097] Neighborhood pattern stability S(c) refers to the concentration of land use types within the neighborhood of a conflict unit c, reflecting whether the unit is nested within a dominant land use pattern. The calculation steps are as follows:

[0098] Neighborhood definition: A 3rd order neighborhood N(c) is constructed with the conflicting cell c as the center, that is, extending 3 grid cells in the upward, downward, left and right and diagonal directions, covering a total of 7×7=49 cells.

[0099] Dominant type statistics: Count the number of rasters for each land type in N(c), denote the dominant type as T_max, and the frequency of occurrence as n_max.

[0100] Stability calculation: S(c) is the ratio of the number of dominant type rasters to the total number of neighboring rasters. The value of S(c) ranges from 0 to 1. A higher value indicates a more stable pattern that is less susceptible to disturbance.

[0101] For example, if forest occupies 41 grids in the neighborhood of a conflict unit, then its S(c) = 41 ÷ 49 ≈ 0.84, indicating that the unit is in a high-stability area and its land attributes should not be significantly changed.

[0102] Historical change frequency H(c) is used to quantify the level of land use change activity in conflict units over the past few years. Its calculation method is as follows:

[0103] Change data extraction: Extract land use records V_t(c) for the unit over the past N years (5 years recommended) from the historical land use database, where t represents the year.

[0104] Change frequency calculation: Count the number of land type changes between adjacent years, denoted as h. For example: if the land type change sequence is {agricultural land → agricultural land → industrial land → industrial land → residential land}, then h = 2.

[0105] Frequency normalization processing: The value ranges from 0 to 1. A higher value indicates more frequent land changes and a higher tolerance for disturbance.

[0106] By unifying the above two dimensions into the same judgment framework, a disturbance sensitivity score R(c) is defined to comprehensively evaluate the disturbance susceptibility of conflict units. The calculation formula is as follows: Where μ and ν are the weighting coefficients for neighborhood stability and historical change frequency, respectively, satisfying It is recommended to set μ = 0.6 and ν = 0.4, meaning that consideration of pattern stability is higher than consideration of historical change trends. R(c) ranges from 0 to 1, with higher values ​​indicating greater suitability for land use conversion operations.

[0107] After obtaining R(c), this invention proposes a spatial transformation path L(c) construction method under the principle of minimum disturbance. Its goal is to transform the land use of the current conflicting unit into an acceptable planned use while minimizing disruption to the original spatial structure. The construction steps are as follows:

[0108] Path candidate set definition: for the current land use type List the set of target types that are allowed to be converted, L*. For example, forest land can be converted into ecological green space but not into industrial land.

[0109] Path cost function setting: for each → The Li path assigns a perturbation cost D(c, i), considering the following factors:

[0110] Land grade difference (low cost of converting from high grade to low grade).

[0111] Coordination with surrounding land use (higher consistency with surrounding land use results in lower costs).

[0112] Policy guidance weight (such as assigning lower costs to key encouraged uses).

[0113] Path selection and sorting: Calculate D(c, i) for all legal paths, and select the path with the smallest D as the priority path to form the set of minimum perturbation paths L(c).

[0114] After obtaining the path set L(c), based on the current regional development direction and land carrying capacity target, further screening and sorting are performed to generate the final feasible transformation sequence T(c), which is constructed as follows:

[0115] Planning suitability analysis: For each new use corresponding to a path, the suitability index A(c, i) is calculated by combining factors such as urban development axis, transportation accessibility, and resource supply capacity.

[0116] Conversion cost weighted ranking: Taking into account both path cost D(c, i) and suitability index A(c, i), a comprehensive score F(c, i) is constructed: ; where η and ζ are adjustment factors used to balance suitability and cost (recommended values ​​are η = 0.7 and ζ = 0.3).

[0117] Sort all feasible paths in descending order of F(c, i) to generate a feasibility transformation sequence T(c) = {t1, t2, ...,tm}, where m is the total number of feasible paths. This sequence is used as input for the subsequent optimization model to select the optimal path.

[0118] To achieve the unity of global coordination and local optimization in territorial spatial planning, after generating the feasibility transformation sequence T(c) for each conflict unit, this invention further proposes an intelligent search and optimization method for the potential solution space S of multiple conflict units to obtain the optimal transformation path set T*.

[0119] Let the set of spatial conflict units be C = {c1, c2, ..., cn}, where each cj corresponds to a generated feasible transformation sequence T(cj) = {t1, t2, ..., tm}, and each ti is a candidate land transformation path for cj. Then, combining all T(cj) forms the Cartesian product scheme space S, which is defined as follows: ; where each element sk∈S is a transformation vector of length n, representing the transformation path number selected for each conflict unit cj.

[0120] Since the number of candidate solutions for each T(cj) may be different, S is typically a high-dimensional, irregular, discrete search space with exponential combinatorial growth characteristics. Therefore, intelligent optimization algorithms are needed to efficiently search for it to obtain the optimal path set T*.

[0121] This invention defines the planning objective function F as a set of multi-objective functions to quantify the overall merits of each candidate solution s_k. It includes the following three main types of objectives:

[0122] Maximizing the integrity of ecological protection This objective measures the degree to which the optimization results minimize the impact on ecological control lines (such as basic farmland, ecological red lines, and water source protection areas). Specifically, it uses the area A_eco(s) of ecologically sensitive units that are intervened as the indicator, with the objective being to minimize: In other words, the optimization direction is to minimize the area of ​​ecological zones that are transformed or developed.

[0123] Maximizing land use efficiency The objective assesses the level of intensive land use after conversion, such as the population or economic output per unit area. A planning suitability coefficient matrix ηjk can be introduced, representing the unit efficiency after converting cj to scheme k. The objective function is: F2(s) = ∑(ηjk) / total area; the objective direction is to maximize efficiency.

[0124] Minimize planning revision costs Considering the disturbance cost Djk for each path ti (e.g., land use change grade difference, construction and supporting costs, resident relocation compensation, etc.), we have: The objective is to minimize cost. The final integrated objective function is represented as a multi-objective vector: .

[0125] To ensure the compliance and feasibility of the optimization results, this invention sets the following two types of constraints:

[0126] Hard constraints The conditions that must be strictly met include:

[0127] Red line constraints: Ecological protection lines, water source protection areas, and basic farmland cannot be illegally converted;

[0128] Legal restrictions: Land designated as non-construction land cannot be converted into construction land;

[0129] Planning intensity threshold: The intensity of a unit land use shall not exceed the target planning limit.

[0130] This type of constraint is implemented by setting a constraint decision goroutine g(s). If the condition is not met, the candidate solution sk is eliminated or the penalty value is set to be extremely large to prevent it from entering the optimization path.

[0131] soft constraints Conditions where a certain degree of deviation is acceptable, but with penalties:

[0132] Neighborhood continuity constraint: The similarity between the new land use and the surrounding land use must exceed a threshold;

[0133] Balanced development intensity: Avoid highly uneven regional development intensity;

[0134] Soft constraints are incorporated into the objective function through a penalty function p(s), influencing the optimization ranking rather than forcibly excluding.

[0135] To find the optimal set of paths T* in a large-scale solution space S, this invention employs a hybrid intelligent search algorithm that combines a multi-objective evolutionary algorithm (represented by NSGA-II) with local search. The specific steps are as follows:

[0136] Each solution sk is represented as a set of integer vectors with a length equal to 1. ;

[0137] Each element is an index of a transformation path in T(cj);

[0138] Initialize N individuals to form a population P0.

[0139] Calculate F1, F2, and F3 for each individual sk;

[0140] Introduce a constraint checker to handle violations. Individual punishments are set;

[0141] Use non-dominated sorting and crowding distance mechanism sorting (NSGA-II strategy).

[0142] Perform single-point crossover and path perturbation mutation on elite individuals to generate offspring. ;

[0143] Offspring and parents are merged into a joint population. .

[0144] Pareto front individuals are selected from R0, and fine-tuning is performed based on the perturbation sensitivity R(c);

[0145] Simulated annealing is used to control the variation temperature and avoid getting trapped in local optima.

[0146] Repeat the iteration for G generations or until the termination criterion is met (e.g., the change in the target value is less than 1%, or there is no improvement for 20 consecutive generations); output the current optimal non-dominated solution set T*.

[0147] Finally, the algorithm outputs the optimal path set T*, where each solution is a combination of transformation paths that satisfies all hard constraints and forms a Pareto optimal set in the objective function F space.

[0148] This invention uses the perturbation sensitivity score R(c) as the core indicator to screen out "characteristic perturbation regions" from all conflict units corresponding to T*. The specific steps are as follows:

[0149] In the previous steps, a disturbance sensitivity score R(c) was constructed based on the neighborhood pattern stability S(c) and the historical change frequency H(c), and its calculation formula is as follows: Where μ and ν are the weighting coefficients for stability and change frequency, respectively, and it is recommended to set μ = 0.6 and ν = 0.4.

[0150] Set a perturbation sensitivity threshold σ (recommended value is 0.65) to filter out conflicting units with R(c) ≥ σ as characteristic perturbation units.

[0151] Spatial clustering algorithms (such as DBSCAN or grid-based eight-neighbor connectivity clustering methods) are used to merge spatially adjacent or nearest-neighbor highly sensitive units into the same spatial patch. Each clustering result is defined as a weakly sensitive region Z_i, and finally, a set of weakly sensitive regions Z = {Z1, Z2, ..., Z_k} is formed.

[0152] To accurately determine the adjustable space and adjustment direction of each weakly sensitive area, this invention proposes to construct a multi-dimensional feature matrix F(Z) to quantify its key attributes such as ecological carrying capacity, structural continuity and traffic accessibility.

[0153] By statistically analyzing the consistency of land use in the surrounding neighborhoods (5th-order neighborhoods) of Zone Z, a structural continuity index C(Z) is defined. The formula is: C(Z) = maximum number of similar land uses ÷ total number of neighborhood grids; the range is between 0 and 1, and the higher the value, the higher the consistency of the surrounding pattern.

[0154] By combining factors such as slope, infrastructure density, and distance to water sources, the average resource carrying capacity within area Z is calculated using a weighted average. Specifically, normalized indices η1 (topographic suitability), η2 (infrastructure supply), and η3 (water resource availability) can be used, and weighted for calculation. A lower E(Z) indicates that area Z is not suitable for high-intensity development and should be prioritized as an ecological buffer or flexible land use.

[0155] Based on road network data, the average shortest path distance from Zone Z to the nearest main road or rail transit node is calculated and converted into a scoring index (the shorter the distance, the higher the score), which can be used to assess its transportation support capacity as a functional land use.

[0156] Finally, C(Z), E(Z), and T(Z) are merged to form the multidimensional feature matrix of the weakly sensitive region: F(Z) = [C(Z), E(Z), T(Z)].

[0157] For the identified weakly sensitive region Zi, this invention designs the following three fine-tuning strategies, the specific strategy to be selected is determined by the numerical structure of F(Zi):

[0158] When C(Z) ≥ 0.7 and E(Z) ≤ 0.4, it indicates that Zone Z is stable but has insufficient carrying capacity. It is recommended to downgrade its planning use level, for example, from Class II construction land to ecological green space or flexible development zone.

[0159] When C(Z) < 0.6 and T(Z) ≥ 0.6, it indicates that the area has a complex layout but good transportation. Its overall efficiency and flexible utilization rate can be improved by nesting multiple land use types within the Z area (such as mixed land use mode).

[0160] When E(Z) is at a critical level (e.g., 0.5 ± 0.1) and C(Z) is high, a flexible boundary strategy is adopted: a priority construction zone and a reversible zone are delineated within the Z zone to ensure room for future adjustments and avoid rigid implementation.

[0161] To ensure that the Z region after fine-tuning still meets the global optimal objective, this invention introduces a secondary optimization verification mechanism to ensure that local changes will not cause adverse disturbances to the overall layout.

[0162] For the adjusted Z-region scheme, the global objective function F(s) is re-substituted, and the changes ΔF of F1, F2, and F3 are evaluated. If the decrease of any objective exceeds 5%, the iterative fine-tuning process begins.

[0163] During the iteration process, simulated annealing or local hill climbing algorithms are used to perturb and optimize the schemes within the Z region, ensuring that a new local optimum is found while maintaining a balance between ecological control and development efficiency.

[0164] Finally, the adjusted land use allocation scheme for Zone Z was finalized and incorporated into the national spatial planning layer P_final as a key control unit. Its sensitivity, adjustment path, and strategy type were recorded in the metadata.

[0165] Example 2, please refer to Figure 2 As shown in this embodiment, the intelligent optimization system for land spatial planning based on multi-source data fusion includes:

[0166] Multi-source spatial data acquisition module: Acquires multi-source spatial data set D within the target planning area;

[0167] Weighted modeling module: Based on the spatial resolution, timeliness and coordinate consistency parameters in the dataset D, construct a multi-source data consistency evaluation model R, and output the multi-source data fusion confidence weight matrix M;

[0168] Conflict identification module: Based on the weight matrix M, perform local sensitivity analysis on the target area to identify a set of spatial conflict units C, wherein the spatial conflict unit C is a grid unit that falls into at least two types of high-priority land use or ecological red lines at the same time.

[0169] Path generation module: For each conflict unit, based on the stability of its neighborhood pattern and the frequency of historical changes, construct the minimum disturbance spatial transformation path L(c) and generate a feasible transformation sequence T(c);

[0170] Optimal path set construction module: Based on the planning objective function and constraints, intelligent search and optimization are performed on the potential solution space S composed of all T(c) combinations to obtain the optimal transformation path set T*, where the planning objective function includes the objectives of maximizing ecological protection integrity, maximizing land use efficiency and minimizing planning correction costs;

[0171] Dynamic adjustment module: Marks the feature disturbance areas in the optimal path set T* as key planning weakly sensitive areas, and performs adjustment strategies on area Z, including dynamic weight adjustment, neighborhood reclassification and flexible land use reservation, and outputs the optimized land spatial layout results.

[0172] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for intelligent optimization of land spatial planning based on multi-source data fusion, characterized by: include: Obtain the multi-source spatial data set D within the target planning area; Based on the spatial resolution, timeliness, and coordinate consistency parameters in the dataset D, a multi-source data consistency evaluation model R is constructed, and the multi-source data fusion confidence weight matrix M is output. Based on the weight matrix M, a local sensitivity analysis is performed on the target region to identify a set of spatial conflict units; For each conflict unit, based on the stability of its neighborhood pattern and the frequency of historical changes, a minimum disturbance spatial transformation path L(c) is constructed, and a feasible transformation sequence T(c) is generated. Based on the planning objective function and constraints, intelligent search and optimization are performed on the potential solution space S composed of all T(c) combinations to obtain the optimal transformation path set T*, where the planning objective function includes the objectives of maximizing ecological protection integrity, maximizing land use efficiency and minimizing planning correction costs. The feature perturbation areas in the optimal path set T* are marked as key planning weakly sensitive areas, and adjustment strategies are implemented for area Z, including dynamic weight adjustment, neighborhood reclassification and flexible land use reservation, and the optimized land spatial layout results are output.

2. The intelligent optimization method for land spatial planning based on multi-source data fusion according to claim 1, characterized in that: The multi-source spatial data set D includes terrain slope data, current land use data, vector boundaries of ecologically sensitive areas, transportation network structure, historical planning boundaries, and construction land approval records.

3. The intelligent optimization method for land spatial planning based on multi-source data fusion according to claim 1, characterized in that: The construction of the multi-source data consistency evaluation model R includes: For each data layer in the multi-source dataset D, extract its original spatial resolution parameter δ_i, data acquisition timestamp τ_i, and projection coordinate system information π_i to form a set of original metadata {δ_i, τ_i, π_i}, where i is the layer number; Based on the proportional relationship between the spatial resolution parameter δ_i and the target uniform resolution δ0, the spatial accuracy deviation factor α_i is calculated. Based on the time difference Δτ_i between the data acquisition timestamp τ_i and the current time τ_now, an exponential decay function is used. Construct a timeliness weighting factor β_i, where k is an empirical coefficient; Based on the consistency relationship between the projected coordinate system π_i and the unified target coordinate system π0, a hash comparison method is used to determine whether they are completely consistent. If they are consistent, γ_i = 1 is assigned. If reprojection transformation is required, γ_i∈(0,1) is calculated based on the boundary offset rate between the transformed layer and the original layer to form a coordinate consistency weight factor. Based on factors α_i, β_i, and γ_i, a consistency comprehensive evaluation model R_i is constructed, and its expression is: Where λ1, λ2, and λ3 are normalized weight coefficients, satisfying ; After normalizing the R_i values ​​of all layers, a confidence weight matrix M={w1, w2, ..., w_n} is formed, where .

4. The intelligent optimization method for land spatial planning based on multi-source data fusion according to claim 3, characterized in that: The identified set of spatial conflict units includes: Each data layer in the weight matrix M is superimposed according to the raster cells, and the comprehensive sensitivity index W(x, y) of each cell is calculated. Set a spatial conflict identification threshold θ. When W(x, y) exceeds θ and there are two or more high-priority conflict attributes, the unit is marked as a candidate conflict unit. Based on the eight-neighborhood analysis method, candidate conflict units are clustered by connectivity, and scattered units with an area smaller than a set area threshold are removed. Output the set of spatially conflicting cells C after filtering.

5. The intelligent optimization method for land spatial planning based on multi-source data fusion according to claim 4, characterized in that: The construction of the minimum perturbation spatial transformation path L(c) and the generation of a feasible transformation sequence T(c) include: Based on the spatial location of conflict unit c, the spatial structure type within its third-order neighborhood is extracted, and the neighborhood pattern stability S(c) is calculated, which is defined as the proportion of the dominant land type within the neighborhood. Query the historical land use records of conflict unit c and its neighborhood, count the number of land type changes in the past N years, calculate the historical change frequency H(c), and construct the disturbance sensitivity score R(c) based on S(c) and H(c); Based on R(c), conflict units are classified into acceptable disturbance levels, and the set of transformation paths L(c) under the principle of minimum disturbance is defined as the set of paths from the current land use status to other permitted planning statuses. The path step size is controlled by planning intensity and ecological constraints. Based on current policy orientation and regional planning objectives, the feasibility of L(c) is screened and ranked to generate a feasibility transformation sequence T(c), in which the paths in T(c) are ranked according to the principle of minimum comprehensive cost.

6. The intelligent optimization method for land spatial planning based on multi-source data fusion according to claim 5, characterized in that: The process of obtaining the optimal conversion path set T* includes: Construct a potential solution space S. Combine the feasibility transformation sequence T(c) of each conflict unit c into a Cartesian combination according to the sequence number to form the potential solution space S. Let each solution be sk, which contains the transformation decision vector of the whole set C. Define the mathematical representation of the planning objective function F and the constraints G. F is a set of multi-objective functions, including maximizing the integrity of ecological protection, maximizing land use efficiency, and minimizing the planning correction cost. It is represented by a weighted linear or Pareto dominance scheme. G includes hard constraints and soft constraints, and is embedded in the optimization process as a feasibility criterion. A hybrid intelligent search algorithm is used to optimize S: a hybrid strategy combining multi-objective evolutionary algorithm and local search is preferred. In each generation of computation, the candidate schemes are constrained, the penalty function is evaluated, and the multi-objective ranking is performed. The applicable encoding is binary or integer vector. The termination criteria are set as the maximum number of generations, the objective convergence threshold, or the number of iterations without improvement. Output the optimal transformation path set T*: Select several paths from the final Pareto front or weighted optimal solution according to the decision-maker's preference to form T*, and calculate the detailed cost, suitability and implementation priority for each selected path.

7. The intelligent optimization method for land spatial planning based on multi-source data fusion according to claim 6, characterized in that: The step of marking the feature perturbation regions in the optimal path set T* as critical planning weakly sensitive regions and implementing adjustment strategies for region Z includes: Based on the disturbance sensitivity score R(c) of each conflict unit in the optimal path set T*, regions with sensitivity scores higher than the set threshold σ are selected, and continuous high-sensitivity units are aggregated into feature disturbance regions by spatial clustering algorithm, which are marked as key planning weakly sensitive regions Z. For each weakly sensitive region Z, calculate the neighborhood structure continuity index and resource carrying capacity index, and construct the multidimensional feature matrix F(Z) of the weakly sensitive region. Based on the F(Z) results and regional planning objectives, a dynamic weight adjustment strategy is adopted to adjust the land use function of area Z, including reallocating land use grades, adjusting land use ratios, or reserving flexible land use boundaries. Iterative verification is performed on the adjusted Z region. The impact of the adjustment scheme on the global planning objective function F is evaluated using simulated annealing or multi-objective local search algorithms. The final adjustment scheme is then output to the planning result layer.

8. A land spatial planning intelligent optimization system based on multi-source data fusion, used to implement the land spatial planning intelligent optimization method based on multi-source data fusion as described in any one of claims 1-7, characterized in that: include: Multi-source spatial data acquisition module: Acquires multi-source spatial data set D within the target planning area; Weighted modeling module: Based on the spatial resolution, timeliness and coordinate consistency parameters in the dataset D, construct a multi-source data consistency evaluation model R, and output the multi-source data fusion confidence weight matrix M; Conflict identification module: Based on the weight matrix M, it performs local sensitivity analysis on the target area and identifies the set of spatial conflict units C; Path generation module: For each conflict unit, based on the stability of its neighborhood pattern and the frequency of historical changes, construct the minimum disturbance spatial transformation path L(c) and generate a feasible transformation sequence T(c); Optimal path set construction module: Based on the planning objective function and constraints, intelligent search and optimization are performed on the potential solution space S composed of all T(c) combinations to obtain the optimal transformation path set T*, where the planning objective function includes the objectives of maximizing ecological protection integrity, maximizing land use efficiency and minimizing planning correction costs; Dynamic adjustment module: Marks the feature disturbance areas in the optimal path set T* as key planning weakly sensitive areas, and performs adjustment strategies on area Z, including dynamic weight adjustment, neighborhood reclassification and flexible land use reservation, and outputs the optimized land spatial layout results.

Images