A city resilience spatial combination pattern recognition method, system and device based on spatial big data mining

Abstract

The application discloses a kind of urban resilience spatial combination mode identification method based on space big data mining, belong to wisdom city and urban space planning technical field.This method obtains the spatial element vector data of research area and carries out grid division, extract the spatial element and its characteristics in each grid unit, quantization topological, direction, distance and so on configuration relationship;Using "element-feature-relation" triple to carry out standardization coding, construct spatial transaction dataset;Based on the improved FP-Growth algorithm, potential characteristic factors are mined, and the output typical spatial combination mode characteristic factor set is output by hierarchical merging, spatial continuity correction and spatial instance verification.The system includes data preprocessing, relationship quantization, coding, pattern mining, post-processing and output module.The application can improve the objectivity, reproducibility and interpretability of identification, and provide quantitative support for typhoon disaster prevention, resilience assessment and spatial planning decision.

Description

Technical Field

[0001] This invention belongs to the field of smart city and urban spatial planning technology, specifically relating to a system and equipment for identifying urban resilient spatial combination patterns based on spatial big data mining. Background Technology

[0002] As my country's urbanization process enters a new stage characterized by intensive development and quality improvement, smart cities, resilient cities, and urban renewal have become key national development directions. Against this backdrop, how to objectively identify stable and synergistic spatial combination patterns that have been tested in long-term practice from massive, multi-dimensional, and heterogeneous urban spatial data has become a critical technical problem that urgently needs to be solved in urban spatial planning and smart city construction.

[0003] Existing technical solutions for urban spatial combination pattern recognition mainly include the following categories:

[0004] The first category is manual analysis methods based on morphological description and typological classification. These methods involve morphological measurements of spatial elements such as buildings, roads, and green spaces, combined with the researcher's experience to classify and summarize patterns. Typical examples include urban fabric analysis and architectural typology analysis. While the methodology of this type is relatively mature, the identification process heavily relies on the researcher's personal experience and subjective judgment. Different researchers may yield significantly different results for the same spatial object, making it difficult to establish objective and reproducible technical standards. Furthermore, this type of method typically only processes a small number of samples, making it unsuitable for analyzing the massive spatial data demands of the urban big data era.

[0005] The second category is data analysis methods based on clustering of single morphological indicators. These methods extract morphological indicators such as building density, street accessibility, and green space ratio, and employ general clustering algorithms such as K-Means and hierarchical clustering to classify and analyze urban space. While this type of method introduces quantitative analysis, it has two significant shortcomings: first, it only focuses on the morphological characteristics of individual elements, failing to depict the structural relationships between different spatial elements, which are precisely the core connotation of spatial combination patterns; second, the output of general clustering algorithms is the grouping result of spatial units, rather than specific and interpretable spatial organization rules, making it difficult to directly translate into planning and design guidelines.

[0006] The third category is automatic analysis methods based on deep learning image recognition. These methods take satellite remote sensing imagery or city maps as input and automatically identify specific spatial patterns using deep learning models such as convolutional neural networks. The limitations of this type of method are: the model is a black box structure, resulting in a lack of interpretability in the output; the training process requires a large number of labeled samples, leading to high labeling costs; and the identified objects are limited by the pixel features of the image, making it difficult to characterize multidimensional configurational relationships based on vector spatial elements.

[0007] The fourth category is association rule discovery methods based on frequent pattern mining. Frequent pattern mining algorithms (such as Apriori and FP-Growth) were originally widely used in commercial transaction data analysis to discover association rules among items in a shopping basket. In recent years, some scholars have also attempted to introduce them into spatial analysis, but existing applications have the following problems: First, a standardized coding structure adapted to spatial data has not been established, making it difficult to reconstruct the mined frequent itemsets into patterns with spatial semantics; second, only the co-occurrence relationship of elements is considered, without incorporating spatial configuration relationships such as topology, orientation, and distance into the mining process; third, there is a lack of post-processing for the mining results, resulting in a large number of redundant, fragmented, and spatially uncorrelated "pseudo-patterns" mixed in the output results, seriously affecting the usability of the recognition results.

[0008] In summary, existing methods suffer from the following specific problems: First, they lack objective and reproducible techniques for identifying urban spatial patterns, relying excessively on human experience and judgment. Second, they lack the ability to quantitatively characterize the multidimensional structural relationships between spatial elements, failing to capture the core structural features of spatial patterns. Third, they lack data mining algorithm improvements adapted to the characteristics of spatial data, and directly applying general data mining algorithms leads to semantic gaps and severe redundancy in the results. Fourth, they lack effective post-processing mechanisms for the mining results, making it difficult to output stable, refined, and spatially relevant typical feature factors. Fifth, the identification results are difficult to link with environmental stress gradients, failing to effectively support the needs of resilient city construction and disaster prevention and mitigation planning.

[0009] This invention is a solution to the above-mentioned problems. Summary of the Invention

[0010] To address the shortcomings of existing urban spatial combination pattern recognition methods, such as strong subjectivity, lack of multi-dimensional configuration relationship characterization, lack of improved mining algorithms adapted to spatial data, and lack of effective post-processing mechanisms, this invention provides a method and system for identifying urban resilience spatial combination patterns based on spatial big data mining. This invention deeply integrates frequent pattern mining algorithms with geospatial analysis. Through multi-dimensional configuration relationship quantification, standardized data structure encoding of "element-feature-relationship" triples, an improved FP-Growth mining algorithm, and a three-stage post-processing workflow, it achieves objective identification of typical spatial combination patterns with stability, synergy, and semantic interpretability from massive spatial data. This provides scientific and reproducible quantitative support for smart city construction, urban resilience assessment, land spatial planning, and the protection and inheritance of historical and cultural heritage.

[0011] To achieve the above-mentioned objectives, the technical solution adopted by this invention is as follows:

[0012] Step (1): Obtain spatial element vector data of the study area through the geographic information system interface, and divide the study area into regular grid units according to the preset grid size. The spatial element vector data includes spatial elements such as buildings, roads, vegetation, water systems, and coastlines, as well as their geometric and attribute information; the grid size is adaptively selected according to the spatial scale of the analysis object.

[0013] Step (2): For each grid cell, the spatial elements and their features contained therein are extracted using a spatial analysis algorithm, and the configuration relationships between each pair of spatial elements are quantified from three dimensions: topology, direction, and distance. Among them, the topological relationship is determined based on the nine-intersection topology model to determine the categories of spatial elements such as disjoint, connected, overlapping, covering, containing, equal, covered, and contained; the directional relationship is obtained by calculating the azimuth angle between the line connecting the centroids of the minimum bounding rectangles of two spatial elements and the positive direction of the horizontal X-axis; the distance relationship uses Euclidean distance or Manhattan distance to quantify the spatial interval.

[0014] Step (3) involves performing discretization encoding on the feature and configuration relationship, and organizing the itemset of each grid cell according to the standardized data structure of "feature-feature-relationship" triples. Specifically, each itemset is represented as:

[0015] The standardized data structure of the element-feature-relation triplet in step S3 is formally represented as follows: Itemset ; in, E Representing spatial elements, A Indicates the characteristics of the elements, R Indicates the configurational relationship between elements, subscript i, j, k, l For numbering, suffix Discretized classification labels are used. For continuous feature values, the natural discontinuity method is used to determine the grading threshold, dividing them into several grading intervals with internal similarity and inter-group differences, and mapping each grading interval to a finite category label. For non-continuous feature values, they are directly mapped to category labels based on their original category attributes. For morphological relationships, topological relationships, and other relationships that inherently have discrete category attributes, they are directly mapped to relationship category labels. For continuous or quasi-continuous relationships such as distance relationships and azimuth relationships, they are discretized and encoded according to the grading intervals determined by the preset grading rules or the natural discontinuity method. After the above processing, the itemsets of each grid cell are organized according to the standardized data structure of "feature-feature-relationship" triples, and the itemsets of all grid cells are aggregated to form a spatial transaction dataset.

[0016] Step (4) involves executing the improved FP-Growth frequent pattern mining algorithm on the spatial transaction dataset. This algorithm avoids the candidate set generation and multiple database scans required by the traditional Apriori algorithm by constructing a Frequent Pattern Tree, significantly improving mining efficiency. Further, this invention requires: using combinations of feature characteristics as antecedents of association rules, and combinations of configurational relationships between features as consequents of association rules; each frequent itemset must simultaneously contain at least two different types of spatial features, corresponding feature characteristics, and at least one configurational relationship; and filtering is performed according to support thresholds (preferably 0.05), confidence thresholds (preferably 0.6), and lift thresholds (preferably 1.5) to extract frequent itemsets that meet the strong association conditions as potential feature factors.

[0017] Step (5) Perform a three-stage post-processing procedure on the latent feature factors:

[0018] The first stage is hierarchical merging, which identifies feature factor pairs that have a strict inclusion relationship. If the feature factors of the parent set meet the threshold condition, the subset is merged into the parent set to eliminate information redundancy.

[0019] The second stage is spatial continuity correction, which identifies continuous spatial elements that are artificially separated at the boundaries of adjacent grids, attempts to merge cross-grid feature factors, and replaces them after verifying their frequency of occurrence, so as to solve the spatial fragmentation problem caused by grid division.

[0020] The third stage is spatial instance verification. Combining satellite remote sensing imagery, high-precision electronic maps, or field survey data, we search for actual spatial scenes that match each feature factor within the study area, eliminate pseudo-patterns that lack real spatial semantic support, and ensure that the output results are interpretable and have practical application value.

[0021] Step (6) outputs the set of typical feature factors after post-processing as the identification result of the urban resilience spatial combination pattern, supporting vector format output and geographic information system visualization.

[0022] Compared with the prior art, the present invention has at least the following beneficial effects:

[0023] First, the spatial configuration relationships of the three dimensions of topology, direction, and distance are systematically incorporated into the mining process, breaking through the limitation of traditional frequent pattern mining that only considers the co-occurrence relationship of elements, and realizing a complete characterization of the core structural features of spatial combination patterns.

[0024] Second, it pioneered the standardized data structure of "element-feature-relationship" triples, establishing a semantically consistent bridging interface for spatial data and frequent pattern mining algorithms. This ensures that the frequent itemsets output by mining can be directly restored to combined patterns with spatial semantics, significantly improving the interpretability of the recognition results.

[0025] Third, the FP-Growth algorithm is improved to adapt to the characteristics of spatial data. By defining the association rule structure where the antecedent is a combination of feature elements and the consequent is a combination of configuration relationships, the mining results directly correspond to the core question of spatial combination patterns: "when spatial elements with specific characteristics exist, what kind of spatial relationship do they tend to form?" This greatly enhances the planning application value of the results.

[0026] Fourth, it features a unique three-stage post-processing workflow: hierarchical merging to eliminate redundancy, spatial continuity correction to resolve grid fragmentation, and spatial instance verification to eliminate pseudo-patterns. The three work together to ensure that the typical feature factors of the final output are not only statistically significant and spatially continuous, but also verifiable in practice, thus making up for the lack of effective post-processing in existing frequent pattern mining methods.

[0027] Fifth, by introducing an optional extension step of environmental stress gradient partitioning data, the method of this invention supports differentiated mining of spatial combination patterns under different stress levels, enabling the method of this invention to directly support the practical needs of key national development directions such as resilient city construction, disaster prevention and mitigation planning, and climate-adaptive design.

[0028] Sixth, the entire process is executed by computer, key parameters can be flexibly configured, supporting the identification needs of different spatial scales, different research areas, and different types of spatial elements. It can also process massive spatial data through performance optimization methods such as spatial indexing, and has good universality, scalability and engineering usability. Attached Figure Description

[0029] Figure 1 This is a flowchart illustrating the overall process of an embodiment of the present invention.

[0030] Figure 2 This is a schematic diagram of the basic topological relationships based on the nine-intersection topology model in an embodiment of the present invention.

[0031] Figure 3 This is a schematic diagram illustrating the principle of calculating the azimuth angle between elements in an embodiment of the present invention.

[0032] Figure 4 This is a schematic diagram of the standardized data structure and discretization encoding process of the "element-feature-relationship" triplet in an embodiment of the present invention.

[0033] Figure 5 This is a flowchart of the improved FP-Growth frequent pattern mining algorithm in an embodiment of the present invention.

[0034] Figure 6 This is a flowchart of the three-stage post-processing in an embodiment of the present invention, which includes three stages: hierarchical merging, spatial continuity correction, and spatial instance verification.

[0035] Figure 7 This is a diagram showing the feature factor identification results of an embodiment of the present invention, taking the explicit characteristics of "mountain and sea landscape" in Xiamen City as an example.

[0036] Figure 8 This is a schematic diagram illustrating the differential mining method based on environmental stress gradients in an embodiment of the present invention.

[0037] Figure 9 This is a system structure block diagram according to an embodiment of the present invention. Detailed Implementation

[0038] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of the application will be further described in detail below with reference to the accompanying drawings. The described embodiments are only a part of the embodiments involved in this invention. All non-innovative embodiments based on this invention by other researchers in the art are within the protection scope of this invention.

[0039] Example 1: Taking Xiamen, a typical typhoon-prone city on the southeastern coast of my country, as an example, this study identifies typical spatial combination patterns in its coastal space. Xiamen boasts a rich mountain and sea landscape, a winding coastline, and dense traditional buildings and modern arcade streets, making it an ideal research object for verifying the method of this invention. Figure 1 As shown, the specific methods include:

[0040] Step (1): Obtain research data and divide it into grids.

[0041] Spatial vector data of Xiamen's coastal area was obtained from the National Geographic Information Public Service Platform and Open Street Map, including elements such as coastline, major mountains, bays, vegetation distribution, water systems, buildings, and road networks. To address the different analytical needs of Xiamen's various visible features, three grid sizes were used: a 4 km × 4 km grid for macro-scale visible feature analysis such as "Mountain and Sea Landscape"; a 2 km × 2 km grid for meso-scale visible feature analysis such as "Bay Necklace" and "Garden Landscape"; and a 500 m × 500 m grid for micro-scale visible feature analysis such as "Warm Homeland".

[0042] Step (2), configuration relationship quantification.

[0043] Taking the analysis of the visible features of the "Mountain and Sea Landscape" within a 4km x 4km grid as an example, the main spatial elements involved in this visible feature include mountains, streams, bays, and built-up areas. For each grid cell, the following configuration relationship quantification is performed:

[0044] Topological relation quantification: based on the nine-intersection model (e.g.) Figure 2As shown), the topological relationships between elements such as mountain-built area, stream-built area, and mountain-bay are determined. For example, in a certain grid cell, the nine-intersection matrix of the mountain and the built area is [[1,0,0], [1,1,0], [1,1,1]], which is determined to be a "connected" relationship;

[0045] Directional relationship quantification: Calculate the azimuth angle between the line connecting the centroids of the smallest circumscribed rectangles of the two elements and the positive direction of the horizontal X-axis (e.g., Figure 3 (As shown). For example, in a certain grid cell, the azimuth angle between the centroid of the mountain and the centroid of the built-up area is 135°, indicating that the built-up area is located southeast of the mountain;

[0046] Distance relationship quantification: Calculate the Euclidean distance between the smallest bounding rectangles of two elements. For example, in a certain grid cell, the closest distance from the mountain boundary to the built-up area boundary is 800 meters.

[0047] Step (3), data structure encoding.

[0048] like Figure 4 As shown in this embodiment, for continuous or quasi-continuous indicators such as average elevation, bay tortuosity, building density, green space ratio, and distance values, the natural discontinuity method is preferentially used to determine the grading thresholds. The natural discontinuity method aims to reduce the differences within each gradation and increase the differences between different gradations. It determines the breakpoints based on the distribution characteristics of the sample data itself, thereby avoiding the weakening of the non-uniform distribution characteristics of spatial data by equidistant grading. For non-continuous indicators such as building type, road grade, vegetation type, and topological relationship category, they are directly mapped to discrete labels according to their category attributes. If directional relationships are expressed in azimuth angle form, they can be encoded using the natural discontinuity method, quadrant partitioning, or preset azimuth intervals as needed for the research.

[0049] Discretization encoding is performed on element features and configuration relationships. Taking "Mountain and Sea Landscape" as an example:

[0050] Mountain features (Average elevation) is discretized into 4 categories according to [0-100m, 100-300m, 300-500m, >500m], and labeled as follows. to ;

[0051] Gulf features (Tortuosity) is discretized into 4 categories according to [1.0-1.5, 1.5-2.0, 2.0-3.0, >3.0], and labeled as follows. to ;

[0052] Features of built-up areas (Building density) is discretized into 4 categories according to [0-0.2, 0.2-0.4, 0.4-0.6, >0.6], and labeled as follows. to ;

[0053] Distance Relationship They are discretized into four categories based on [<500m, 500-1000m, 1000-2000m, >2000m], and labeled as follows: to ;

[0054] Directional relationship Based on [0°-90°, 90°-180°, 180°-270°, 270°-360°], they are discretized into 4 categories, labeled as follows: to ;

[0055] Topological relationships The eight categories determined directly using the nine intersection criteria are labeled as R3_1 to R3_8.

[0056] An example of an itemset for a grid cell is:

[0057] ;

[0058] This indicates that within this grid cell, there exists a mountain of moderate height (300-500m). A low-density (0-0.2) built-up area The distance between the two is between 500-1000m, and the built-up area is located to the northeast of the mountain. The itemsets of all grid cells are aggregated to form a spatial transaction dataset.

[0059] Step (4), Improved FP-Growth Frequent Pattern Mining.

[0060] like Figure 5 As shown, the FP-Growth algorithm is executed on the spatial transaction dataset. The minimum support threshold is set to 0.05, the minimum confidence threshold to 0.6, and the minimum lift threshold to 1.5. The algorithm first scans the dataset to count the frequency of each item and sorts them in descending order, removing items with a support threshold below the threshold; secondly, it constructs a frequent pattern tree and merges records by sharing prefix paths; finally, it recursively mines frequent itemsets.

[0061] To meet the semantic requirements of spatial composition patterns, this invention filters the output of the standard FP-Growth algorithm, retaining only frequent itemsets that simultaneously satisfy the following conditions: (a) they contain at least two different types of spatial features; (b) they contain at least one feature of the corresponding feature; (c) they contain at least one configurational relationship between the two types of features; and (d) the antecedent is a combination of feature combinations, and the consequent is a combination of configurational relationships between features. After mining and filtering, a set of latent feature factors is obtained.

[0062] Taking "Mountain and Sea Landscape" as an example, a typical latent characteristic factor discovered is: The itemset has a support of 0.08, a confidence level of 0.72, and a lift of 1.95, indicating that the spatial combination of "medium-height mountains and low-density built-up areas at a distance of 500-1000m, with the built-up areas located northeast of the mountains" is highly prevalent and strongly correlated in the study area.

[0063] Step (5), three-stage post-processing (such as...) Figure 6 (As shown).

[0064] The first stage is hierarchical merging. This assumes the simultaneous existence of the following three latent feature factors that satisfy a threshold:

[0065] Feature factors ;

[0066] Feature factors ;

[0067] Feature factors ;

[0068] Where A and B are both proper subsets of C. Determine whether C meets the threshold requirements (support 0.06, confidence 0.68, lift 1.82). If it does, merge A and B into C, delete the entries of A and B, and retain C as the feature factor of the merged set.

[0069] The second stage involves spatial continuity correction. Imagine a long coastline... Divided into two adjacent grids, grid Characteristic factors formed in , grid Characteristic factors formed in The present invention detected exist and The boundary exhibits spatial continuity; an attempt is made to merge two eigenfactors into a cross-grid eigenfactor. The frequency of occurrence of the feature in the study area is verified. If the frequency meets the support threshold (e.g., 0.06), the original two feature factors are replaced with the merged feature factor.

[0070] The third stage is spatial instance verification. For each feature factor processed in the first two stages, matching instances are searched in the study area using a geographic information system combined with satellite imagery. Taking feature factor C as an example, multiple real spatial instances are found, such as the Gulangyu Sunlight Rock-Bijia Mountain-Xiamen Island built-up area and the Yunding Rock-Qianpu built-up area, verifying that the feature factor has sufficient actual spatial semantic support and thus is retained. Conversely, if a feature factor has fewer than a preset threshold of matching instances in the entire study area (e.g., 3), it is discarded.

[0071] After three stages of post-processing, the final set of typical characteristic factors under the explicit traits of Xiamen's "mountain and sea landscape" was obtained, such as... Figure 7 As shown. Similarly, the above process was performed on other explicit traits such as "bay necklace", "garden landscape" and "warm home" to obtain the corresponding set of typical feature factors, resulting in dozens of typical spatial combination patterns with stability, synergy and interpretability.

[0072] Step (6), output the result.

[0073] All typical characteristic factors are output in two forms: structured data tables and geographic information system vector files. The visualization module generates spatial distribution maps, instance distribution maps, and typical spatial scene diagrams for each characteristic factor, which helps planning decision-makers to intuitively understand and apply the identification results.

[0074] Example 2: Differential mining combined with environmental stress gradient.

[0075] This embodiment illustrates how, under optional extension steps, the present invention can be combined with environmental stress gradient partitioning data to identify resilient spatial combination patterns with environmental adaptability orientation (such as...). Figure 8 (As shown).

[0076] Taking the typhoon adaptability study of Xiamen City as an example, the cumulative impact gradient zoning data of typhoons from external input is introduced (which can be obtained by another technology of the inventors associated with this invention, namely the gradient zoning method based on the cumulative impact of historical typhoons). This data divides Xiamen City into high, medium and low typhoon cumulative impact zones.

[0077] The spatial transaction dataset was divided into a high-stress subset and a low-stress subset based on the cumulative impact gradient of typhoons. FP-Growth mining (step (4)) was performed on each subset to obtain the sets of high-stress and low-stress feature factors. Through set difference analysis, the following feature factors with typhoon adaptability were identified:

[0078] Feature factors (This represents "a highly tortuous bay and a medium-sized windbreak forest appearing together within a distance of less than 500m").

[0079] The support of this characteristic factor is 0.12 in areas with high cumulative impact and only 0.02 in areas with low cumulative impact, showing a significant difference. This indicates that the spatial combination pattern is a typical adaptive pattern formed by Xiamen City under long-term typhoon environment selection, reflecting the local wisdom in disaster prevention and construction.

[0080] Example 3: Application in the protection of historical and cultural cities.

[0081] Besides being applicable to typhoon adaptability research in coastal cities, this invention can also be extended to the field of historical and cultural city preservation. Taking the study of traditional settlements in a historical and cultural city as an example:

[0082] Obtain vector data of spatial elements such as historical buildings, streets and alleys, water systems, ancestral halls, and feng shui forests of the famous city, and divide them into a fine grid of 200 meters × 200 meters;

[0083] The configurational relationships between the above elements are quantified from three dimensions: topology, orientation, and distance.

[0084] Spatial transaction datasets are formed by encoding using the "element-feature-relationship" triplet.

[0085] Perform FP-Growth mining to extract typical spatial combination patterns such as "ancestral hall located in the center of settlement, main street unfolding along feng shui axis, and water system flowing around the south side of settlement";

[0086] After three stages of post-processing and spatial instance verification, a set of typical characteristic factors of the traditional settlements in this historic city is output, providing data support for the protection and inheritance of historical culture, digital archiving of traditional settlements, and preservation of distinctive features in renovation and upgrading.

[0087] Example 4: Integration and application in a smart city platform.

[0088] The system of this invention can be integrated as a spatial big data analysis module into a smart city integrated management platform or a city information model (CIM) platform, providing the following functions:

[0089] Real-time access to multi-source urban spatial data, and regular updates to spatial transaction datasets;

[0090] It supports planning decision-makers in customizing parameters such as research scope, grid size, and key thresholds to quickly uncover spatial combination patterns under specific regions and themes.

[0091] It provides objective pattern recognition support for urban health checks, urban renewal diagnosis, planning of key areas, and renewal of historical blocks;

[0092] It achieves data interoperability with disaster prevention and mitigation command systems, land and space planning information systems, and cultural relic protection and management systems.

[0093] Alternative implementation methods:

[0094] For the selection of grid size in step (1), in addition to 500m, 2km, and 4km in the embodiment, other sizes such as 100m, 1km, and 10km can also be used, or irregular divisions (such as administrative units, street units, and natural village units) can be used as transaction processing units.

[0095] For the quantification of configuration relationships in step (2), in addition to the nine-intersection model, the simplified RCC-8 (region connectivity calculation) model or RCC-5 model can also be used; in addition to azimuth, the MBR (minimum bounding rectangle) directional relationship model or cone directional model can also be used; in addition to Euclidean distance, Manhattan distance, shortest path distance, travel time distance, etc. can also be used for distance relationships.

[0096] For the mining algorithm in step (4), in addition to FP-Growth, it can also be replaced by other frequent pattern mining algorithms such as Apriori, Eclat, and H-Mine to adapt to different data scales and computing resource constraints.

[0097] For spatial instance verification in step (5), in addition to manual verification, deep learning target detection models (such as YOLO and MaskR-CNN) can be used to automatically identify spatial scenes in satellite remote sensing images that match the feature factors, thus achieving full automation of the process.

[0098] For the environmental stress gradient in step (8), in addition to the cumulative impact of typhoons, other environmental stress data such as flood risk index, earthquake intensity distribution, urban heat island intensity, and air pollution concentration can also be used, so that the method of the present invention can be applied to various types of urban resilience research scenarios.

[0099] System Implementation Example: Based on the above method, this invention also provides a system for recognizing urban resilience spatial combination patterns based on spatial big data mining. The system structure is as follows: Figure 9 As shown, it includes the following modules:

[0100] The spatial data preprocessing module is responsible for acquiring spatial element vector data through the geographic information system interface, performing preprocessing such as coordinate system 1, topology checking, and attribute field standardization, and performing gridding according to the grid size specified by the user.

[0101] The configuration relationship quantification module encapsulates the nine-intersection topology determination algorithm, azimuth calculation algorithm, and Euclidean / Manhattan distance calculation algorithm, which automatically quantifies the configuration relationship between spatial elements from three dimensions: topology, direction, and distance.

[0102] The data standardization module implements the discretization encoding of feature and configuration relationship, organizes itemsets according to the standardized data structure of "feature-feature-relationship", and summarizes them to form a spatial transaction dataset;

[0103] The frequent pattern mining module encapsulates the improved FP-Growth algorithm, supports user-defined support, confidence, and lift thresholds, performs frequent itemset mining and semantic filtering on spatial transaction datasets, and outputs latent feature factors.

[0104] The three-stage post-processing module implements three sub-functions: hierarchical merging, spatial continuity correction, and spatial instance verification, respectively, to refine and control the quality of potential feature factors; the spatial instance verification sub-module can call satellite remote sensing image services or schedule field survey and record data.

[0105] The results output and visualization module outputs typical feature factor sets in various formats such as structured tables, Shapefiles, and GeoJSON, and generates visualization outputs such as spatial distribution maps, instance distribution maps, and typical scene diagrams.

[0106] The application interface module provides a standard RESTful API interface to connect with upper-level applications such as smart city integrated management platforms, city information modeling (CIM) platforms, land and space planning information systems, urban resilience assessment systems, and historical and cultural protection management systems.

[0107] The aforementioned modules can be deployed on standalone servers, workstations, or cloud computing platforms, supporting various modes such as single-machine operation, distributed computing, and microservice deployment. To address the demands of processing massive spatial data, this system integrates spatial index structures such as R-trees and quadtrees into the front end of the frequent pattern mining module, significantly reducing the algorithm's time complexity. Experiments show that compared to the unoptimized version, the computation time can be reduced by more than 70% when processing spatial transaction datasets with millions of itemsets.

[0108] Computer Device Embodiment: The computer device embodiment of the method of the present invention includes a memory and a processor. The memory stores a computer program implementing the method of the present invention, and the processor, by calling the program, executes the steps of the method described in any one of claims 1 to 11. The computer device may be a general-purpose server, workstation, personal computer, or a virtual instance deployed on a cloud computing platform. The memory includes, but is not limited to, non-volatile storage devices such as hard disks, solid-state drives, optical disks, and cloud storage media. The computer-readable storage medium embodiment of the method of the present invention is a non-volatile storage medium storing the aforementioned computer program, which, when executed by the processor, implements all the steps of the method of the present invention.

[0109] The above description is only a preferred embodiment of the present invention. For those skilled in the art, several modifications and substitutions can be made without departing from the principles and concepts of the present invention, and all equivalent modifications and substitutions should fall within the protection scope of the present invention.

Claims

1. A method for identifying urban resilience spatial combination patterns based on spatial big data mining, characterized in that, Includes the following steps: S1. Obtain spatial element vector data of the study area through the geographic information system interface, and divide the study area into regular grid units according to the preset grid size; S2. For each grid cell, the spatial elements and their features contained therein are extracted by spatial analysis algorithms, and the configuration relationship between each pair of the spatial elements is quantified from three dimensions: topological relationship, directional relationship and distance relationship, to generate spatial configuration feature data of the grid cell. S3. Perform discretization encoding on the feature and configuration relationship, organize the itemset of each grid cell according to the standardized data structure of feature-relation triplet, and summarize the itemsets of all grid cells in the study area to form a spatial transaction dataset. S4. Execute the improved FP-Growth frequent pattern mining algorithm on the spatial transaction dataset, using the combination of spatial element features as the antecedent of the association rule and the combination of configuration relationships between spatial elements as the consequent of the association rule, and extract frequent itemsets that satisfy strong association based on preset support threshold, confidence threshold and lift threshold, as potential spatial combination pattern feature factors. S5. Perform a three-stage post-processing process on the potential spatial combination pattern feature factors. The post-processing includes hierarchical merging, spatial continuity correction and spatial instance verification, and outputs a set of typical spatial combination pattern feature factors. S6. The set of characteristic factors of the typical spatial combination pattern is used as the identification result of the urban resilience spatial combination pattern, which is used to guide the construction of smart cities, spatial planning decisions and urban resilience assessment.

2. The method for identifying urban resilience spatial combination patterns based on spatial big data mining according to claim 1, characterized in that, The quantification of topological relationships in step S2 is based on the nine-intersection topological model. By calculating the intersection state matrix of the interior, boundary, and exterior of two spatial elements, one of the following topological relationship categories is obtained: disjoint, connected, overlapping, covering, containing, equal, covered, and contained. The quantification of directional relationships is obtained by calculating the azimuth angle between the line connecting the centroids of the minimum bounding rectangles of the two spatial elements and the positive direction of the horizontal X-axis. The azimuth angle ranges from 0° to 360°. The quantification of distance relationships uses Euclidean distance or Manhattan distance to measure the spatial interval between the two spatial elements.

3. The method for identifying urban resilience spatial combination patterns based on spatial big data mining according to claim 1, characterized in that, The standardized data structure of the element-feature-relation triplet in step S3 is formally represented as follows: Itemset ; in, E Representing spatial elements, A Indicates the characteristics of the elements, R Indicates the configurational relationship between elements, subscript i, j, k, l For numbering, suffix For discretization classification labels; The discretization encoding process includes: determining the grading threshold for continuous feature values ​​using the natural discontinuity method, and mapping the continuous feature values ​​to a finite number of categories according to the grading threshold; directly mapping non-continuous feature values ​​to category labels; discretizing and encoding continuous configuration relationship values ​​according to preset grading rules or the natural discontinuity method, and directly mapping non-continuous configuration relationships to relationship category labels.

4. The method for identifying urban resilience spatial combination patterns based on spatial big data mining according to claim 1, characterized in that, In step S4, the support threshold ranges from 0.03 to 0.10, the confidence threshold ranges from 0.5 to 0.8, and the lift threshold ranges from 1.2 to 2.

0. The latent spatial combination pattern feature factors must satisfy the following: the itemset of each feature factor simultaneously contains at least two different types of spatial elements, corresponding feature features, and at least one configuration relationship between the elements.

5. The method for identifying urban resilience spatial combination patterns based on spatial big data mining according to claim 1, characterized in that, The hierarchical merging in step S5 includes the following sub-steps: S5.1.1 Traverse the set of latent feature factors and identify feature factor pairs that have a strict inclusion relationship, that is, the itemset of one feature factor is a proper subset of the itemset of another feature factor. S5.1.2 Determine whether the parent set feature factors simultaneously meet the requirements of the support threshold, confidence threshold, and lift threshold; S5.1.3 If the parent set satisfies the condition, the feature factors of the subset are merged into the parent set, and the subset entries are deleted; if the parent set does not satisfy the condition, both subset and parent set entries are retained. The spatial continuity correction includes the following sub-steps: S5.2.1 Detect spatial elements with spatial continuity at the boundaries of adjacent grid cells; S5.2.2 Identify different characteristic factors involving continuous spatial elements in adjacent grid cells; S5.2.3, Try to merge different feature factors into feature factors that cross grid boundaries, and verify the actual occurrence frequency of the merged feature factors within the study area; S5.2.4 If the frequency of occurrence of the merged feature factor meets the support threshold, then the merged feature factor replaces the original feature factor that was artificially divided by the grid boundary. The spatial instance verification uses a geographic information system combined with satellite remote sensing imagery, high-precision electronic maps, or field survey data to search for actual spatial scenes that match each feature factor within the study area. When a feature factor cannot find a matching instance or the number of matching instances is lower than a preset instance number threshold, it is determined that the feature factor lacks real spatial semantic support and is removed from the set of typical feature factors.

6. The method for identifying urban resilience spatial combination patterns based on spatial big data mining according to claim 1, characterized in that, The method also includes an environmental stress response identification step: S7.1 Based on the environmental stress gradient partitioning data from external input, the spatial transaction dataset is divided into subsets of different stress levels; S7.

2. Perform the frequent pattern mining step S4 on each subset of datasets to obtain the feature factor set under each stress level. S7.

3. Through set difference analysis, identify characteristic factors that are significantly present in high stress areas and absent or significantly different in low stress areas, as resilient spatial combination patterns with environmental adaptability orientation.

7. The method for identifying urban resilience spatial combination patterns based on spatial big data mining according to claim 6, characterized in that, The environmental stress gradient zoning data includes one or more combinations of typhoon cumulative impact gradient zoning, flood risk gradient zoning, earthquake risk gradient zoning, and urban heat island intensity gradient zoning.

8. The method for identifying urban resilience spatial combination patterns based on spatial big data mining according to claim 1, characterized in that, The method further includes a performance optimization step: before step S4 is executed, the spatial transaction dataset is preprocessed based on a spatial index structure, wherein the spatial index structure includes one or more combinations of R-tree, quadtree, and grid index.

9. A system for recognizing urban resilience spatial combination patterns based on spatial big data mining, characterized in that, include: The spatial data preprocessing module is used to acquire spatial element vector data and perform gridding. The configuration relationship quantification module is used to quantify the configuration relationships between spatial elements from three dimensions: topological relationship, directional relationship, and distance relationship. The data standardization module is used to discretize and encode feature characteristics and configuration relationships according to the triplet standardized data structure and organize them into itemsets. The frequent pattern mining module is used to extract latent feature factors that meet threshold conditions based on the improved FP-Growth algorithm. The three-stage post-processing module is used to perform hierarchical merging, spatial continuity correction, and spatial instance verification. The results output and visualization module is used to output the set of characteristic factors of typical spatial combination patterns and support the visualization display of geographic information systems. The application interface module is used to connect to smart city platforms, spatial planning information systems, and urban resilience assessment systems.

10. A computer device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when loaded onto the processor, implements the method according to any one of claims 1 to 8.

Images