A land change degree analysis method and system based on spatiotemporal data

Abstract

The application belongs to the technical field of land change degree analysis methods, and particularly relates to a land change degree analysis method and system based on space-time data, which comprises the following steps: acquiring multi-temporal remote sensing data of a space monitoring unit and extracting time sequence characteristics; constructing a preliminary time similarity pair set based on the similarity between time sequence characteristic analysis units; constructing a space connection relationship graph by using a graph theory method, and identifying a candidate change group; performing spatial clustering analysis on the candidate group, and screening a group with strong internal connectivity; performing time sequence consistency analysis on the screened group, and fusing interannual change characteristics to generate a space-time comprehensive homogeneity score to determine a homogenous change region; identifying a dominant change mode sequence of each homogenous region and classifying the evolution law thereof; and finally performing grouping optimization based on the evolution law type. The application solves the problem of time-space fragmentation and difficulty in accurately identifying homogenous change regions in traditional methods, and realizes high-precision automatic identification and law interpretation of change regions.

Description

Technical Field

[0001] This invention belongs to the technical field of land change degree analysis methods, specifically relating to a land change degree analysis method and system based on spatiotemporal data. Background Technology

[0002] In the field of land use change analysis, current methods primarily rely on independent calculations of change intensity or statistical analysis of type transition frequencies for individual plots or monitoring units. This approach typically treats each spatial unit as an isolated analytical object, identifying changes in land cover type or physical indicators by comparing remote sensing imagery data from different temporal phases. While such methods provide basic change information, their analytical paradigm has significant limitations: they primarily focus on the longitudinal change trajectory of the unit itself, generally neglecting the intrinsic correlations between different land units along their temporal evolution paths, as well as their spatial proximity or functional coupling relationships. This approach fragments regions that should exhibit coordinated change characteristics into discrete individuals, making it difficult to reveal regional evolutionary patterns at a macroscopic level. Specifically, there is currently a lack of computational methods that can simultaneously characterize temporal evolution similarity and spatial correlation strength, and provide a unified measurement of both. Summary of the Invention

[0003] The technical problem to be solved by this invention is to overcome the problems of high computational complexity, limited performance, and poor interactive experience when performing real-time, interactive semantic segmentation of large-scale 3D point clouds in a web browser environment.

[0004] To address the aforementioned technical problems, a first aspect of the present invention discloses a method for analyzing the degree of land change based on spatiotemporal data, the method comprising the following steps:

[0005] S1. Acquire multi-temporal remote sensing data from multiple spatial monitoring units and extract the temporal characteristics of each unit;

[0006] S2. Based on the temporal characteristics, analyze the temporal similarity between the spatial monitoring units, and select unit pairs with temporal similarity higher than a first preset threshold to form a preliminary set of temporal similarity pairs;

[0007] S3. Based on the preliminary time similarity pair set, construct a connection relationship diagram between spatial monitoring units, and identify connected regions with a number of units greater than a preset scale threshold from the diagram as candidate change groups;

[0008] S4. Perform spatial clustering analysis on the candidate change groups, and filter out groups whose internal spatial connectivity is higher than the second preset threshold to form a set of change groups after preliminary screening.

[0009] S5. Perform time-series consistency analysis on each group in the preliminary set of change groups, and integrate their interannual variation characteristics to generate a spatiotemporal comprehensive homogeneity score; if the score is higher than the third preset threshold, the group is determined to be a homogeneous change region.

[0010] S6. Identify the dominant change pattern sequence for each homogeneous change region, and classify the evolution pattern type of the region based on the regularity of the sequence;

[0011] S7. Based on the aforementioned evolutionary pattern type, all homogeneous change regions are grouped and optimized to obtain the final change region grouping results.

[0012] As an optional implementation, in the first aspect of the present invention, step S1 includes:

[0013] The study area is divided into regular grids using a spatial grid division method, with each grid serving as a spatial monitoring unit.

[0014] Acquire multi-temporal remote sensing images of each space monitoring unit, and perform geometric registration, radiometric calibration and atmospheric correction to generate a surface reflectance image sequence;

[0015] From the surface reflectance image sequence, pixel-level spectral reflectance time-series curves are extracted, and at least one spectral index among the normalized vegetation index and water index is calculated, thereby generating an annual seasonal fluctuation sequence of spectral indexes.

[0016] The pixel-level spectral reflectance time series curve is combined with the annual seasonal fluctuation sequence of the spectral index to obtain a complete time series feature set at the unit level.

[0017] As an optional implementation, in the first aspect of the present invention, step S2 includes:

[0018] Calculate the Euclidean distance of temporal shape similarity between each pair of spatial monitoring units and generate a temporal similarity distance matrix;

[0019] The unit pairs whose Euclidean distance is lower than the fourth preset threshold are marked as preliminary time similarity pairs to form a preliminary set;

[0020] Extract the key inflection points of the time series of unit pairs in the preliminary set, and use the dynamic time warping algorithm to calculate the matching distance between the inflection points;

[0021] Unit pairs whose matching distance is lower than the fifth preset threshold are retained to form the final time similarity pair set, which serves as the initial time similarity pair set.

[0022] As an optional implementation, in the first aspect of the present invention, step S3 includes:

[0023] A spatial connection graph is constructed using spatial monitoring units as nodes and unit pairs in the preliminary time similarity pair set as edges;

[0024] All connected components are extracted from the graph using a depth-first search algorithm;

[0025] Connected components containing more than a preset size threshold are identified as candidate variation groups, wherein the preset size threshold is dynamically set based on the ratio of the total area of ​​the study area to the average area of ​​the units.

[0026] As an optional implementation, in the first aspect of the present invention, step S4 includes:

[0027] Obtain the annual seasonal fluctuation series of the boundary coordinates and spectral indices of the internal units of the candidate change groups;

[0028] A density-based spatial clustering algorithm is used to perform spatial neighborhood expansion clustering on the units within the group, where the neighborhood radius is adaptively adjusted according to the average geographical distance of the units.

[0029] If the spatial clusters formed by clustering have internal connectivity higher than the second preset threshold, the candidate change group is retained; otherwise, it is discarded, resulting in a set of change groups after preliminary screening.

[0030] As an optional implementation, in the first aspect of the present invention, step S5 includes:

[0031] Extract the time-series spectral reflectance curves of all units within the variation group, and perform time-series shape clustering using dynamic time warping distance or K-means algorithm to obtain multiple time-series sub-patterns and their representative units;

[0032] Based on the representative unit of the time-series sub-pattern, the overall profile coefficient of the group is calculated as a time-series consistency index;

[0033] Calculate the slope value of the interannual gradient evolution characteristics of the annual seasonal fluctuation series of spectral indicators within the group;

[0034] The contour coefficients and slope values ​​are weighted and fused to generate a group spatiotemporal comprehensive homogeneity score;

[0035] Groups with scores higher than a third preset threshold are identified as homogeneous change regions.

[0036] As an optional implementation, in the first aspect of the present invention, step S6 includes:

[0037] Based on the order of occurrence and transformation relationship of representative units of time-series sub-modes within homogeneous change regions, an annual summary of dominant change pattern sequences is generated.

[0038] Analyze the repetition frequency and periodic stability variance of the main changing directions in the sequence summary;

[0039] Based on the rule that the repetition frequency is higher than the sixth preset threshold and the periodic stability variance is lower than the preset variance threshold, the evolutionary patterns are classified into sudden, gradual, cyclic, or mixed types.

[0040] As an optional implementation, in the first aspect of the present invention, step S7 includes:

[0041] A cohesive hierarchical clustering algorithm is used to cluster homogeneous change regions based on the type of evolutionary pattern.

[0042] Calculate the cosine similarity or Jaccard index between branches of the tree structure during clustering;

[0043] The regions contained in branches with a similarity higher than the fourth preset threshold are merged to obtain the optimized grouping result.

[0044] A second aspect of this invention discloses a land change analysis system based on spatiotemporal data, used to implement the land change analysis method based on spatiotemporal data described in any of the above embodiments, the system comprising:

[0045] The data acquisition module is used to acquire multi-temporal remote sensing data from multiple spatial monitoring units and extract the temporal characteristics of each unit.

[0046] The analysis module is used to analyze the temporal similarity between the spatial monitoring units based on the temporal characteristics, and to filter out unit pairs with temporal similarity higher than a first preset threshold to form a preliminary set of temporally similar pairs.

[0047] The generation module is used to construct a connection graph between spatial monitoring units based on the preliminary time similarity pair set, and to identify connected regions with a number of units greater than a preset scale threshold from the graph as candidate change groups;

[0048] The filtering module is used to perform spatial clustering analysis on the candidate change groups, filter out groups whose internal spatial connectivity is higher than a second preset threshold, and form a set of change groups after preliminary filtering.

[0049] The determination module is used to perform temporal consistency analysis on each group in the initially screened set of change groups, and integrate their interannual variation characteristics to generate a spatiotemporal comprehensive homogeneity score; if the score is higher than a third preset threshold, the group is determined to be a homogeneous change region.

[0050] The identification module is used to identify the dominant change pattern sequence of each homogeneous change region and classify the evolution pattern type of the region based on the regularity of the sequence;

[0051] The grouping optimization module is used to optimize the grouping of all homogeneous change regions based on the evolutionary pattern type, and obtain the final change region grouping result.

[0052] A third aspect of this invention discloses another land change analysis system based on spatiotemporal data, the system comprising:

[0053] Memory containing executable program code;

[0054] A processor coupled to the memory;

[0055] The processor calls the executable program code stored in the memory to execute a land change analysis method based on spatiotemporal data disclosed in the first aspect of the present invention.

[0056] The fourth aspect of the present invention discloses a computer-readable storage medium storing computer instructions, which, when invoked by a processor, are used to execute a land change analysis method based on spatiotemporal data disclosed in the first aspect of the present invention.

[0057] Compared with the prior art, the beneficial effects of the present invention are:

[0058] By acquiring multi-temporal remote sensing data and extracting unit temporal features, a data foundation is provided for analysis. By analyzing temporal similarity and constructing spatial connectivity diagrams, the automatic discovery of potential homogeneous regions is achieved, overcoming the limitations of spatiotemporal fragmentation. Spatial clustering analysis verifies the spatial entityness of groups. By generating spatiotemporal comprehensive homogeneity scores through temporal consistency analysis and feature fusion, accurate and objective determination of homogeneous change regions is achieved. By generating change pattern sequence summaries and classifying evolutionary patterns, the depth of understanding of change patterns is enhanced. Finally, group optimization based on evolutionary patterns ensures the scientific validity and practicality of the final zoning results, improving the effectiveness of land change analysis in decision support. This invention, through a series of interconnected processes, deeply integrates temporal similarity analysis and spatial correlation analysis, achieving automatic and accurate identification and pattern interpretation of homogeneous change regions. Attached Figure Description

[0059] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings, wherein:

[0060] Figure 1 This is a flowchart illustrating a method for analyzing the degree of land change based on spatiotemporal data, as disclosed in an embodiment of the present invention.

[0061] Figure 2 This is a schematic diagram of the structure of a land change analysis system based on spatiotemporal data disclosed in an embodiment of the present invention;

[0062] Figure 3 This is a schematic diagram of another land change analysis system based on spatiotemporal data disclosed in an embodiment of the present invention. Detailed Implementation

[0063] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and 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.

[0064] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or end that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or ends.

[0065] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0066] This invention discloses a method and system for analyzing land change based on spatiotemporal data. By acquiring multi-temporal remote sensing data and extracting unit temporal features, a data foundation is provided for analysis. By analyzing temporal similarity and constructing a spatial connectivity graph, the automatic discovery of potential homogeneous regions is achieved, overcoming the limitations of spatiotemporal fragmentation. Spatial clustering analysis verifies the spatial entityness of groups. By generating a spatiotemporal comprehensive homogeneity score through temporal consistency analysis and feature fusion, accurate and objective determination of homogeneous change regions is achieved. By generating change pattern sequence summaries and classifying evolutionary patterns, the depth of understanding of change patterns is improved. Finally, group optimization based on evolutionary patterns ensures the scientific validity and practicality of the final zoning results, enhancing the effectiveness of land change analysis in decision support.

[0067] Example 1

[0068] Please see Figure 1 , Figure 1 This is a flowchart illustrating a method for analyzing land change based on spatiotemporal data, as disclosed in an embodiment of the present invention. Figure 1 The described method for analyzing land change based on spatiotemporal data is applied to a data processing chip, processing terminal, or processing server, which can be a local server or a cloud server; this embodiment of the invention does not limit the application of such a method. Figure 1 As shown, this method for analyzing the degree of land change based on spatiotemporal data may include the following operations:

[0069] S1. Acquire multi-temporal remote sensing data from multiple spatial monitoring units and extract the temporal characteristics of each unit.

[0070] Specifically, the study area is divided into multiple regular grids using a pre-defined spatial grid division method, with each grid serving as an independent analysis unit. Subsequently, multi-temporal remote sensing image sequences covering a relatively long time span are acquired from a professional remote sensing data platform for each spatial monitoring unit. The acquired raw image data undergoes rigorous geometric registration and cropping based on the unit's geographic coordinates to ensure accurate geographical correspondence between images from different temporal phases. On this basis, radiometric calibration and atmospheric correction are performed pixel-by-pixel on the images to eliminate atmospheric interference, resulting in a surface reflectance image sequence that truly reflects the physical properties of the land surface. The extraction of temporal features includes two core parts: first, extracting the multispectral reflectance values ​​of each pixel from the corrected image sequence and arranging them chronologically to form a pixel-level spectral reflectance temporal curve; second, based on this reflectance data, further calculating spectral indicators sensitive to changes in specific land features, such as the normalized difference vegetation index (NDVI) and water index, and organizing them by year and season to generate an annual seasonal fluctuation sequence of spectral indicators. Finally, the above features of all pixels within each spatial monitoring unit are spatially aggregated to form a complete unit-level temporal series feature set.

[0071] As can be seen, the above-mentioned systematic data acquisition and processing process has laid a high-precision, multi-dimensional foundation for subsequent analysis of time series features. This invention not only achieves continuous and refined data capture of surface change processes, but more importantly, the extracted pixel-level spectral reflectance time series curves and annual seasonal fluctuation sequences of spectral indicators together constitute a comprehensive time series feature that can simultaneously characterize subtle changes in surface physical properties and periodic fluctuations of specific land features. This provides sufficient data support for accurately analyzing the temporal evolution similarity between spatial units, fundamentally overcoming the limitations of insufficient data utilization and single feature dimensions in traditional methods, and ensuring the reliability of subsequent change group identification and evolution law analysis.

[0072] S2. Based on the temporal characteristics, analyze the temporal similarity between the spatial monitoring units, and select unit pairs with temporal similarity higher than a first preset threshold to form a preliminary set of temporally similar pairs.

[0073] Specifically, based on the complete time-series feature set of each spatial monitoring unit, the Euclidean distance calculation method is used to quantify the degree of difference in the overall shape of the time-series feature vectors of different units, generating a time-series similarity distance matrix. The system compares the distance values ​​in the matrix with a first preset threshold, filtering out unit pairs with distance values ​​below the threshold to form a preliminary time-similarity pair set. To further improve the accuracy of the judgment, the system extracts the key inflection points of the time series for each unit pair in this set and uses a dynamic time warping algorithm to calculate the matching path length between the inflection point sequences. By comparing the matching distance with a second preset threshold, the system finally retains those unit pairs that have a high degree of consistency in both overall shape and key change nodes, forming a reliable preliminary time-similarity pair set.

[0074] As can be seen, through the aforementioned dual screening mechanism, the system achieves an objective and accurate assessment of the similarity of temporal evolution patterns among spatial monitoring units. This method not only identifies units similar in overall change trends but also effectively eliminates misjudgments caused by minor timeline shifts or noise interference through matching verification at key turning points. This provides high-quality, high-confidence foundational data for subsequently constructing spatial connectivity maps that truly reflect spatiotemporal relationships, fundamentally overcoming the problem of misjudgments of correlations that may arise from relying solely on a single similarity index in traditional methods, and ensuring the accuracy of input for subsequent change group identification processes.

[0075] S3. Based on the preliminary time similarity pair set, construct a connection relationship diagram between spatial monitoring units, and identify connected regions with a number of units greater than a preset scale threshold from the diagram as candidate change groups.

[0076] Specifically, each spatial monitoring unit is abstracted as a node in a graph structure, and each pair of units in the initial set of temporally similar pairs is used as an edge connecting the corresponding nodes to construct a spatial connectivity graph. The system uses a depth-first search algorithm to traverse the graph, identifying all interconnected subgraphs, i.e., connected components. For each connected component, the system counts the number of nodes it contains and compares this value with a preset scale threshold. If the number of units in a connected component is greater than the threshold, it indicates that it has sufficient statistical significance and spatial scale, and the system identifies it as a candidate change group.

[0077] As can be seen, through the graph theory construction and connectivity analysis methods described above, the system transforms unit pair associations based on temporal similarity into spatial group relationships with a clear topological structure. This invention can systematically identify potential change regions from massive discrete units that may be spatially continuous or adjacent and have intrinsic temporal correlations. The introduction of a preset scale threshold effectively filters out small, isolated connected components generated by accidental similarity, ensuring that the discovered candidate change groups possess physical significance and statistical stability for regional analysis, providing a structured set of candidate objects for subsequent spatial clustering analysis and homogeneity determination.

[0078] S4. Perform spatial clustering analysis on the candidate change groups, and filter out groups whose internal spatial connectivity is higher than the second preset threshold to form a set of change groups after preliminary screening.

[0079] Specifically, the system acquires the boundary coordinates of each candidate change group and the annual seasonal fluctuation sequences of spectral indicators for all spatial monitoring units within it. A density-based spatial clustering algorithm is used as the grouping algorithm to perform spatial neighborhood expansion clustering analysis on the units within the candidate change groups. This algorithm aggregates spatially adjacent and similar units into the same sub-cluster based on the geographical proximity and spectral similarity between units. After clustering, the system evaluates the internal connectivity of each formed spatial cluster, specifically by calculating the proportion of core units within the cluster or the tightness of connections between units. This calculated value is compared with a second preset threshold. If it is higher than this threshold, the cluster is determined to be a valid spatial cluster with strong internal connectivity, and its corresponding candidate change group is retained; otherwise, it is discarded. Through this process, a preliminary set of selected change groups is formed.

[0080] As can be seen, through the aforementioned spatial clustering analysis and connectivity threshold judgment, the system has verified the spatial entity nature of the candidate groups initially identified based on temporal similarity. This method can effectively identify those changing regions that truly possess close adjacency relationships and internal structural coherence in spatial distribution. This ensures that the groups ultimately entering subsequent analysis not only have similarities in temporal evolution patterns but also possess the spatial basis for management as unified geographical entities, avoiding the misclassification of spatially discrete and fragmented unit sets as homogeneous regions and improving the practical operability of the identified changing regions in land spatial planning applications.

[0081] S5. Perform time-series consistency analysis on each group in the preliminary screening set of change groups, and integrate their interannual variation characteristics to generate a spatiotemporal comprehensive homogeneity score. If the score is higher than the third preset threshold, the group is determined to be a homogeneous change region.

[0082] Specifically, the system extracts the spectral reflectance time-series curves of all spatial monitoring units within each group in the initially screened set of change groups. Using dynamic time-warped distance or K-means algorithms as grouping algorithms, the system performs secondary clustering analysis on the time series shapes within each group, identifying different change rhythms or patterns, thus obtaining multiple time-series sub-patterns and their representative units. Subsequently, the system calculates the silhouette coefficient of the entire group based on these representative units. This coefficient quantifies temporal consistency by comparing the similarity of each unit with other units within the group and its difference from the nearest external group. Simultaneously, the system calculates the interannual gradient evolution slope value of the annual seasonal fluctuation series of spectral indicators within the group; this slope reflects the strength of the change trend. Finally, the system integrates the silhouette coefficient and interannual gradient slope value into a single spatiotemporal comprehensive homogeneity score using a weighted fusion algorithm.

[0083] As can be seen, by integrating the aforementioned temporal consistency analysis with the quantitative fusion of interannual variation characteristics, the system achieves an objective and comprehensive assessment of the homogeneity within change groups. This method not only identifies the inherent consistency of groups in their temporal evolution rhythm but also considers the intensity characteristics of change trends, thus avoiding the biases that may arise from single-dimensional assessments. The introduction of a spatiotemporal comprehensive homogeneity score provides a unified and quantifiable standard for determining homogeneous change regions, ensuring that the ultimately identified regions possess a high degree of inherent consistency in both temporal and spatial dimensions, providing a reliable foundation for subsequent evolutionary pattern analysis and precise control.

[0084] S6. Identify the dominant change pattern sequence for each homogeneous change region, and classify the evolution pattern type of the region based on the regularity of the sequence.

[0085] Specifically, for each identified homogeneous change region, the system analyzes multiple temporal sub-pattern representative units obtained through time series shape clustering. Based on the order of appearance of these representative units on the time axis and their transformation relationships, the system constructs a pattern sequence that reflects the dominant change process of the region. Based on this sequence, the system extracts key change direction features and uses time series pattern recognition algorithms to analyze the repetition frequency and periodicity intensity of the main change directions in the sequence. Specifically, the repetition frequency is calculated by counting the number of times a specific pattern appears within the observation time window, and the periodic stability of the change pattern is evaluated using a periodic detection algorithm. Finally, based on preset classification rules, such as classifying regions with high repetition frequency and strong periodic stability as cyclical and regions where changes occur concentrated in short periods as sudden, the system completes the automated classification and identification of the evolutionary pattern type of each homogeneous change region.

[0086] As can be seen, through the in-depth mining and regular quantitative analysis of the dominant change pattern sequence, the system has achieved a precise interpretation of the internal evolutionary logic within homogeneous change regions. This method can not only identify the static attribute of "homogeneity" of a region, but also further diagnose the core patterns and regularities of its dynamic evolution, such as whether it exhibits periodic fluctuations or unidirectional gradual development. This provides crucial insights into understanding the intrinsic driving forces of regional change and predicting its future development trends, enabling land change analysis to move from static description of the current situation to dynamic interpretation of mechanisms, ultimately supporting the formulation of precise and differentiated spatial management strategies based on differences in evolutionary patterns.

[0087] S7. Based on the aforementioned evolutionary pattern type, all homogeneous change regions are grouped and optimized to obtain the final change region grouping results.

[0088] Specifically, the system acquires the classification results of evolutionary patterns across all homogeneous change regions. It employs a grouping algorithm, such as agglomerative hierarchical clustering, to perform cluster analysis on homogeneous change regions using evolutionary pattern types as feature vectors. During clustering, the system constructs a tree-like structure and calculates the similarity between branches, specifically quantifying the degree of matching of evolutionary patterns between branches using cosine similarity or the Jaccard index. The system compares the calculated similarity values ​​with a preset threshold. If the branch similarity exceeds the threshold, all homogeneous change regions contained in that branch are merged into the same group. By iteratively executing this merging process, the system ultimately generates optimized grouping results for the change regions.

[0089] As can be seen, through the above-described grouping optimization process based on evolutionary patterns, the system achieves refined and scientific merging of homogeneous change regions. This method ensures that the final grouping results not only reflect the similarity of regions in spatiotemporal characteristics but also fully consider the consistency of their inherent evolutionary dynamics. This makes the grouping results more logically sound and stable, enabling them to more accurately reveal the overall pattern and internal connections of regional changes, and providing a reliable and intuitive basis for regional division for subsequent differentiated management and planning decisions regarding national land space.

[0090] As an optional embodiment, step S1 in the above steps includes:

[0091] The study area is divided into regular grids using a spatial grid division method, with each grid serving as a spatial monitoring unit.

[0092] Acquire multi-temporal remote sensing images of each space monitoring unit, and perform geometric registration, radiometric calibration and atmospheric correction to generate a surface reflectance image sequence;

[0093] From the surface reflectance image sequence, pixel-level spectral reflectance time-series curves are extracted, and at least one spectral index among the normalized vegetation index and water index is calculated, thereby generating an annual seasonal fluctuation sequence of spectral indexes.

[0094] The pixel-level spectral reflectance time series curve is combined with the annual seasonal fluctuation sequence of the spectral index to obtain a complete time series feature set at the unit level.

[0095] In this embodiment of the invention, a spatial grid division method with equal spacing or a custom resolution is used to divide the study area into a regular grid array, with each grid cell serving as an independent spatial monitoring unit. The system acquires multi-temporal remote sensing image sequences for each grid cell from a remote sensing data platform, and sequentially performs geometric registration (eliminating positional bias), radiometric calibration (converting DN values ​​to radiance values), and atmospheric correction (eliminating the effects of atmospheric scattering and absorption) on the original images to generate a surface reflectance image sequence. Subsequently, the system extracts multispectral band reflectance values ​​pixel by pixel and arranges them in chronological order to form a pixel-level spectral reflectance time-series curve. Simultaneously, based on reflectance data, spectral indices such as the Normalized Difference Vegetation Index (NDVI = (NIR-Red) / (NIR+Red)) or the Normalized Difference Water Index (NDWI = (Green-NIR) / (Green+NIR)) are calculated and aggregated by year and season to generate annual seasonal fluctuation sequences. Finally, the pixel-level reflectance time series curves and spectral index fluctuation sequences are aligned by time and combined into a complete time series feature set at the unit level, where NIR is the near-infrared band, NDVI is the Normalized Difference Vegetation Index, and Red is the red light band.

[0096] As can be seen, the above-described systematic data processing workflow enables standardized extraction and multi-dimensional characterization of surface change information. Regular grid partitioning ensures the spatial regularity of the analysis units, multi-step radiometric correction guarantees the accuracy of the data's physical meaning, and the combination of pixel-level time-series curves and spectral index fluctuation sequences simultaneously captures the continuous variation patterns of surface physical properties and the periodic response characteristics of specific land features. This provides a complete feature foundation for subsequent time-series similarity analysis, encompassing both spectral details and temporal dynamics, overcoming the shortcomings of traditional methods that rely on single features or insufficient utilization of temporal information.

[0097] As an optional embodiment, step S2 in the above steps includes:

[0098] Calculate the Euclidean distance of temporal shape similarity between each pair of spatial monitoring units and generate a temporal similarity distance matrix;

[0099] The unit pairs whose Euclidean distance is lower than the fourth preset threshold are marked as preliminary time similarity pairs to form a preliminary set;

[0100] Extract the key inflection points of the time series of unit pairs in the preliminary set, and use the dynamic time warping algorithm to calculate the matching distance between the inflection points;

[0101] Unit pairs whose matching distance is lower than the fifth preset threshold are retained to form the final time similarity pair set, which serves as the initial time similarity pair set.

[0102] In this embodiment of the invention, the Euclidean distance of temporal shape similarity between pairwise spatial monitoring units is calculated. Let the complete time series feature vectors of unit i and unit j be respectively... and (The vector length L represents the feature values ​​at L time points), then the Euclidean distance calculation formula is:

[0103]

[0104] in, This represents the characteristic value of unit i at the k-th time point (such as reflectance or spectral index value in a specific band). Let represent the value corresponding to unit j, and L be the total length of the time series. The system performs this calculation on all unit pairs, generating a time series similarity distance matrix. Subsequently, the values ​​of each unit pair in the matrix are... The value is compared with the fourth preset threshold. Unit pairs that are lower than the threshold are marked as preliminary time similarity pairs, forming a preliminary set.

[0105] For each pair of units in the initial set, the system extracts key inflection points (such as local extremum sequences P and Q) from its time series. The Dynamic Time Warping (DTW) algorithm is used to calculate the matching distance between the inflection point sequences: an m×n cumulative cost matrix D is constructed (m and n are the lengths of sequences P and Q, respectively), where:

[0106]

[0107] Here, The Euclidean distance between the two points is given by the final matching distance. The system compares the matching distance with a fifth preset threshold. Unit pairs that are below this threshold are retained, forming the final set of time-similar pairs.

[0108] As can be seen, through the dual verification mechanism of preliminary screening by Euclidean distance and matching of key turning points of dynamic time warping, the system has achieved multi-level and high-precision evaluation of the similarity of temporal evolution patterns among spatial monitoring units. This invention can not only identify unit pairs that are similar in overall sequence morphology, but also effectively eliminate misjudgments caused by time phase shift or local noise interference through alignment and matching of key change nodes, ensuring that the similarity judgment results have both overall consistency and local key feature matching degree. This provides high-quality and high-reliability basic data for the subsequent construction of a network structure that truly reflects spatiotemporal correlation.

[0109] As an optional embodiment, step S3 in the above steps includes:

[0110] A spatial connection graph is constructed using spatial monitoring units as nodes and unit pairs in the preliminary time similarity pair set as edges;

[0111] All connected components are extracted from the graph using a depth-first search algorithm;

[0112] Connected components containing more than a preset size threshold are identified as candidate variation groups, wherein the preset size threshold is dynamically set based on the ratio of the total area of ​​the study area to the average area of ​​the units.

[0113] In this embodiment of the invention, the first step is to abstract each spatial monitoring unit as a node in a graph structure, and use each pair of units in the initial set of temporally similar pairs as an edge connecting the corresponding nodes, thereby constructing an undirected spatial connection graph. The system then uses a depth-first search (DFS) algorithm to extract all connected components from the graph. The core process of the DFS algorithm is as follows: starting from any unvisited node, recursively visit all its unvisited neighbor nodes until the current connected component has been traversed; repeat this process until all nodes have been visited. The algorithm is implemented by maintaining a visit status array (e.g., visited[i] indicates whether node i has been visited) and a recursive stack, and its time complexity is O(V+E), where V is the number of nodes (i.e., the total number of spatial monitoring units) and E is the number of edges (i.e., the number of initial temporally similar pairs). After traversal, each connected component corresponds to a set of nodes. The system then calculates the number of nodes contained in each connected component and compares it with a preset scale threshold. This threshold is dynamically set according to the ratio of the total area of ​​the study area (S_total) to the average area of ​​the unit (A_unit), and the calculation formula is:

[0114]

[0115] in, This is a preset scaling factor (e.g., 0.001). The total area of ​​the study region (unit: hectares). This represents the average area (in hectares) of a single spatial monitoring unit. If the number of nodes in a connected component is greater than... Then it is identified as a candidate change group.

[0116] As can be seen, through the graph theory construction and connectivity analysis described above, the system transforms unit pair associations based on temporal similarity into spatial group relationships with a clear topological structure. This method can automatically identify potential change regions from massive discrete units that may be spatially continuous or adjacent and have intrinsic temporal correlations. The dynamically set scale threshold effectively filters out small, isolated connected components resulting from accidental similarity, ensuring that the discovered candidate change groups possess physical significance and statistical stability for regional analysis. This provides a structured set of candidate objects for subsequent spatial clustering analysis and homogeneity determination, thus solving the problem of insufficient regional correlation identification caused by the spatiotemporal dimensional fragmentation in traditional methods.

[0117] As an optional embodiment, step S4 in the above steps includes:

[0118] Obtain the annual seasonal fluctuation series of the boundary coordinates and spectral indices of the internal units of the candidate change groups;

[0119] A density-based spatial clustering algorithm is used to perform spatial neighborhood expansion clustering on the units within the group, where the neighborhood radius is adaptively adjusted according to the average geographical distance of the units.

[0120] If the spatial clusters formed by clustering have internal connectivity higher than the second preset threshold, the candidate change group is retained; otherwise, it is discarded, resulting in a set of change groups after preliminary screening.

[0121] In this embodiment of the invention, the boundary coordinates (e.g., the latitude and longitude of the corner points of the smallest bounding rectangle) of each candidate change group and the annual seasonal fluctuation sequence of the spectral indicators of all spatial monitoring units within it are obtained. The system uses a density-based spatial clustering algorithm (e.g., DBSCAN) to perform spatial neighborhood expansion clustering on the units within the group. During the clustering process, the key parameter neighborhood radius (eps) is adaptively adjusted according to the average geographical distance of the units: the geographical distance (e.g., Euclidean distance) between all units in the group is calculated, and the average value is taken as the average geographical distance of the units (D_avg). Then, the neighborhood radius is set to eps = k × D_avg, where k is a preset adjustment coefficient (usually taken as 1.5-2.0). The DBSCAN algorithm is executed through the following steps:

[0122] For each unit point p, calculate its neighborhood N_eps(p) = {q | d(p,q) ≤ eps}, where d(p,q) is the Euclidean distance between points p and q:

[0123] d

[0124] in and These are the geographic coordinates (e.g., longitude and latitude) of points p and q, respectively.

[0125] If the number of points in the neighborhood |N_eps(p)| ≥ minPts (minimum number of points, preset by the user), then p is marked as a core point;

[0126] Spatial clusters are generated by expanding the core point.

[0127] After clustering, the system calculates the internal connectivity score for each spatial cluster:

[0128]

[0129] If the score is higher than the second preset threshold (e.g., 0.7), it is determined that the cluster has strong internal connectivity and the corresponding candidate change group is retained; otherwise, it is eliminated.

[0130] As can be seen, the system effectively screens candidate change groups based on the aforementioned density clustering-based spatial connectivity verification mechanism. This method can effectively distinguish between spatially tightly clustered continuous regions and discretely distributed sets of units, ensuring that the ultimately retained groups have clear geographical boundaries and consistent internal structures. This avoids potential spatial fragmentation misjudgments that may result from relying solely on temporal similarity, provides candidate objects with practical spatial significance for subsequent homogeneity analysis, and improves the operability and reliability of change area identification results in land spatial planning.

[0131] As an optional embodiment, step S5 in the above steps includes:

[0132] Extract the time-series spectral reflectance curves of all units within the variation group, and perform time-series shape clustering using dynamic time warping distance or K-means algorithm to obtain multiple time-series sub-patterns and their representative units;

[0133] Based on the representative unit of the time-series sub-pattern, the overall profile coefficient of the group is calculated as a time-series consistency index;

[0134] Calculate the slope value of the interannual gradient evolution characteristics of the annual seasonal fluctuation series of spectral indicators within the group;

[0135] The contour coefficients and slope values ​​are weighted and fused to generate a group spatiotemporal comprehensive homogeneity score;

[0136] Groups with scores higher than a third preset threshold are identified as homogeneous change regions.

[0137] In this embodiment of the invention, the time-series curves of spectral reflectance of all spatial monitoring units within the change group are extracted. The system uses dynamic time warping distance or K-means algorithm for time series shape clustering. The formula for calculating dynamic time warping distance is as follows: Given two time series X = (x1, x2, ..., x_m) and Y = (y1, y2, ..., y_n), where x_i and y_j represent the feature values ​​(such as reflectance) of the series at time points i and j, respectively, an m×n cumulative cost matrix D is constructed, where D[i,j] = dist(x_i, y_j) + min(D[i-1,j], D[i,j-1], D[i-1,j-1]), and dist(x_i,y_j) is the Euclidean distance between points x_i and y_j (calculated as √(x_i - y_j)²), and the final DTW distance is D[m,n]. The K-means algorithm iteratively optimizes the time series to K clusters to minimize the intra-cluster variance, and after clustering, it obtains multiple time series sub-patterns and their representative units (such as cluster center curves).

[0138] Subsequently, the system calculates the profile coefficient of the entire group as a temporal consistency index based on the representative units of the temporal sub-patterns. For each unit i in the group, a(i) = the average distance between i and other units in the same cluster, and b(i) = the average distance between i and the nearest other cluster unit, then the profile coefficient s(i) of unit i is = (b(i) - a(i)) / max(a(i), b(i)). The overall profile coefficient S_group of the group is the average of all units s(i).

[0139] Simultaneously, the system calculates the slope value of the interannual gradient evolution characteristics of the annual seasonal fluctuation series of spectral indicators within the group. By performing linear regression on the annual series, a fitted model y=ωt+b is obtained, where t is the year, y is the spectral indicator value, and the slope ω is calculated using the least squares method: ω=Σ[(t_i-t_mean)(y_i-y_mean)] / Σ(t_i - t_mean)², where t_mean and y_mean are the average values ​​of the year and the indicator value, respectively.

[0140] Finally, the system performs weighted fusion of the contour coefficient S_group and the absolute slope value |ω| to generate a spatiotemporal homogeneity score H_score=γ×S_group+(1-γ)×|ω|, where γ is a preset weight coefficient (0<γ<1). If H_score is higher than the third preset threshold, the group is determined to be a homogeneous variation region.

[0141] As can be seen, through the aforementioned temporal shape clustering, silhouette coefficient calculation, interannual gradient slope extraction, and weighted fusion, the system achieves an objective quantitative assessment of the homogeneity within change groups. This method can simultaneously consider the consistency of time series shapes and the intensity of change trends, thereby accurately identifying homogeneous change regions with high intrinsic unity in both spatiotemporal dimensions. It avoids the limitations of single-indicator judgment in traditional methods, improves the accuracy and reliability of change region identification, and provides a high-quality foundation for subsequent evolutionary pattern analysis.

[0142] As an optional embodiment, step S6 in the above steps includes:

[0143] Based on the order of occurrence and transformation relationship of representative units of time-series sub-modes within homogeneous change regions, an annual summary of dominant change pattern sequences is generated.

[0144] Analyze the repetition frequency and periodic stability variance of the main changing directions in the sequence summary;

[0145] Based on the rule that the repetition frequency is higher than the sixth preset threshold and the periodic stability variance is lower than the preset variance threshold, the evolutionary patterns are classified into sudden, gradual, cyclic, or mixed types.

[0146] In this embodiment of the invention, for each homogeneous change region, based on the time-series sub-pattern representative units obtained by time series shape clustering within it, the occurrence order and transformation relationship of these units on the time axis are analyzed. Specifically, the system extracts the occurrence frequency of each sub-pattern by year, determines the dominant change pattern of each year (i.e. the sub-pattern with the highest occurrence frequency), and generates a regional annual dominant change pattern sequence summary. This summary is a time series that records the changes of the dominant pattern each year.

[0147] Subsequently, the repetition frequency and periodic stability variance of the main directions of change in the sequence summary were analyzed. The formula for calculating the repetition frequency is:

[0148] F_repeat=C / T

[0149] Where C represents the number of times the main direction of change occurs within the observation period T, and T is the total number of years. The calculation of the periodic stability variance is based on the detected period lengths: assuming the system identifies multiple period lengths L1, L2, ..., L using period detection algorithms (such as autocorrelation function or Fourier analysis). n The formula for calculating the periodic stability variance is:

[0150] Var=Σ(L_i-L_mean)² / n

[0151] Where L_mean is the average cycle length (L_mean=ΣL_i / n), and n is the number of detected cycles.

[0152] Finally, the system classifies regions according to preset rules: if the repetition frequency F_repeat is higher than the sixth preset threshold (e.g., 0.5, indicating the pattern occurs at least once every two years) and the periodic stability variance Var is lower than the preset variance threshold (e.g., 1.0, indicating small fluctuations in period length), the region is classified as cyclical; if F_repeat is low but changes occur concentratedly in short periods, it is classified as bursty; if F_repeat is low and changes show a slow, monotonous trend, it is classified as gradual; those that do not meet the above conditions are classified as mixed. It should be noted that the specific calculation formula is not detailed in the document; this is a refinement based on general time series analysis knowledge, which will not be elaborated upon further here.

[0153] As can be seen, through the above-mentioned sequence summary generation, quantitative analysis of repetition frequency and periodic stability variance, and application of classification rules, the system realizes the automated and objective identification of the evolution patterns of homogeneous change regions. This invention can accurately extract the repetitive and periodic characteristics of change patterns from time series data, thereby distinguishing different evolution types such as sudden, gradual, and cyclical, providing in-depth insights into understanding regional change dynamics and supporting the formulation of differentiated control strategies in territorial spatial planning.

[0154] As an optional embodiment, step S7 in the above steps includes:

[0155] A cohesive hierarchical clustering algorithm is used to cluster homogeneous change regions based on the type of evolutionary pattern.

[0156] Calculate the cosine similarity or Jaccard index between branches of the tree structure during clustering;

[0157] The regions contained in branches with a similarity higher than the fourth preset threshold are merged to obtain the optimized grouping result.

[0158] In this embodiment of the invention, a cohesive hierarchical clustering algorithm is used to cluster homogeneous variation regions. This algorithm uses the evolutionary pattern characteristics of each region as an input vector, initially with each region forming its own cluster. The algorithm iteratively merges the most similar clusters until all regions are merged into one cluster, forming a tree-like structure. Similarity calculation is based on inter-cluster distance: during each merge, the algorithm calculates the inter-cluster distance. Common methods include the average link distance method, which calculates the average distance between all point pairs in two clusters. Assuming cluster A and cluster B contain m and n regions respectively, their average link distance formula is:

[0159]

[0160] in This represents the feature vector distance (e.g., Euclidean distance) between regions a and b. However, in this step, cosine similarity or the Jaccard index is used for similarity evaluation.

[0161] Cosine similarity: For the feature vectors X and Y of two branches (clusters) (such as the frequency vector of evolutionary patterns), the formula is:

[0162]

[0163] Where X·Y is the dot product, and |X| is the vector norm. The range is [-1, 1], and a higher value indicates a stronger similarity.

[0164] Jaccard index: If the branches are considered as a set of regions, the formula is:

[0165]

[0166] Where A and B are the sets of regions contained in the two branches, and |·| represents the size of the set. The value range is [0,1], and the higher the value, the greater the overlap.

[0167] During clustering, the system calculates the similarity between branches (i.e., intermediate clusters) of the tree structure. If the cosine similarity or Jaccard index is higher than a fourth preset threshold (e.g., 0.7), all regions contained in the corresponding branch are merged. Finally, the merging process is controlled by the threshold to obtain the optimized grouping results.

[0168] As can be seen, through the aforementioned hierarchical clustering and branch similarity threshold merging mechanism, the system achieves regularity-driven grouping optimization of homogeneous change regions. This method can automatically adjust the region merging strategy based on the similarity of evolutionary patterns, ensuring that the final grouping results retain the consistency of spatiotemporal evolutionary characteristics while avoiding excessive segmentation or merging. This enhances the logical rationality and stability of change region grouping and provides a scientific basis for differentiated management of national land space.

[0169] Example 2

[0170] Please see Figure 2 , Figure 2 This is a schematic diagram of a land change analysis system based on spatiotemporal data, as disclosed in an embodiment of the present invention. Figure 2 The described land change analysis system based on spatiotemporal data can be applied to data processing chips, processing terminals, or processing servers. The processing server can be a local server or a cloud server; this embodiment of the invention does not impose any limitations. Figure 2 As shown, the land change analysis system based on spatiotemporal data can include the following operations:

[0171] The data acquisition module 201 acquires multi-temporal remote sensing data from multiple spatial monitoring units and extracts the temporal characteristics of each unit.

[0172] Specifically, the study area is divided into multiple regular grids using a pre-defined spatial grid division method, with each grid serving as an independent analysis unit. Subsequently, multi-temporal remote sensing image sequences covering a relatively long time span are acquired from a professional remote sensing data platform for each spatial monitoring unit. The acquired raw image data undergoes rigorous geometric registration and cropping based on the unit's geographic coordinates to ensure accurate geographical correspondence between images from different temporal phases. On this basis, radiometric calibration and atmospheric correction are performed pixel-by-pixel on the images to eliminate atmospheric interference, resulting in a surface reflectance image sequence that truly reflects the physical properties of the land surface. The extraction of temporal features includes two core parts: first, extracting the multispectral reflectance values ​​of each pixel from the corrected image sequence and arranging them chronologically to form a pixel-level spectral reflectance temporal curve; second, based on this reflectance data, further calculating spectral indicators sensitive to changes in specific land features, such as the normalized difference vegetation index (NDVI) and water index, and organizing them by year and season to generate an annual seasonal fluctuation sequence of spectral indicators. Finally, the above features of all pixels within each spatial monitoring unit are spatially aggregated to form a complete unit-level temporal series feature set.

[0173] As can be seen, the above-mentioned systematic data acquisition and processing process has laid a high-precision, multi-dimensional foundation for subsequent analysis of time series features. This invention not only achieves continuous and refined data capture of surface change processes, but more importantly, the extracted pixel-level spectral reflectance time series curves and annual seasonal fluctuation sequences of spectral indicators together constitute a comprehensive time series feature that can simultaneously characterize subtle changes in surface physical properties and periodic fluctuations of specific land features. This provides sufficient data support for accurately analyzing the temporal evolution similarity between spatial units, fundamentally overcoming the limitations of insufficient data utilization and single feature dimensions in traditional methods, and ensuring the reliability of subsequent change group identification and evolution law analysis.

[0174] Analysis module 202 analyzes the temporal similarity between the spatial monitoring units based on the temporal characteristics, and filters out unit pairs with temporal similarity higher than a first preset threshold to form a preliminary set of temporally similar pairs.

[0175] Specifically, based on the complete time-series feature set of each spatial monitoring unit, the Euclidean distance calculation method is used to quantify the degree of difference in the overall shape of the time-series feature vectors of different units, generating a time-series similarity distance matrix. The system compares the distance values ​​in the matrix with a first preset threshold, filtering out unit pairs with distance values ​​below the threshold to form a preliminary time-similarity pair set. To further improve the accuracy of the judgment, the system extracts the key inflection points of the time series for each unit pair in this set and uses a dynamic time warping algorithm to calculate the matching path length between the inflection point sequences. By comparing the matching distance with a second preset threshold, the system finally retains those unit pairs that have a high degree of consistency in both overall shape and key change nodes, forming a reliable preliminary time-similarity pair set.

[0176] As can be seen, through the aforementioned dual screening mechanism, the system achieves an objective and accurate assessment of the similarity of temporal evolution patterns among spatial monitoring units. This method not only identifies units similar in overall change trends but also effectively eliminates misjudgments caused by minor timeline shifts or noise interference through matching verification at key turning points. This provides high-quality, high-confidence foundational data for subsequently constructing spatial connectivity maps that truly reflect spatiotemporal relationships, fundamentally overcoming the problem of misjudgments of correlations that may arise from relying solely on a single similarity index in traditional methods, and ensuring the accuracy of input for subsequent change group identification processes.

[0177] The generation module 203 constructs a connection graph between spatial monitoring units based on the preliminary time similarity pair set, and identifies connected regions with a number of units greater than a preset scale threshold from the graph as candidate change groups.

[0178] Specifically, each spatial monitoring unit is abstracted as a node in a graph structure, and each pair of units in the initial set of temporally similar pairs is used as an edge connecting the corresponding nodes to construct a spatial connectivity graph. The system uses a depth-first search algorithm to traverse the graph, identifying all interconnected subgraphs, i.e., connected components. For each connected component, the system counts the number of nodes it contains and compares this value with a preset scale threshold. If the number of units in a connected component is greater than the threshold, it indicates that it has sufficient statistical significance and spatial scale, and the system identifies it as a candidate change group.

[0179] As can be seen, through the graph theory construction and connectivity analysis methods described above, the system transforms unit pair associations based on temporal similarity into spatial group relationships with a clear topological structure. This invention can systematically identify potential change regions from massive discrete units that may be spatially continuous or adjacent and have intrinsic temporal correlations. The introduction of a preset scale threshold effectively filters out small, isolated connected components generated by accidental similarity, ensuring that the discovered candidate change groups possess physical significance and statistical stability for regional analysis, providing a structured set of candidate objects for subsequent spatial clustering analysis and homogeneity determination.

[0180] The filtering module 204 performs spatial clustering analysis on the candidate change groups, filters out groups whose internal spatial connectivity is higher than a second preset threshold, and forms a set of change groups after preliminary filtering.

[0181] Specifically, the system acquires the boundary coordinates of each candidate change group and the annual seasonal fluctuation sequences of spectral indicators for all spatial monitoring units within it. A density-based spatial clustering algorithm is used as the grouping algorithm to perform spatial neighborhood expansion clustering analysis on the units within the candidate change groups. This algorithm aggregates spatially adjacent and similar units into the same sub-cluster based on the geographical proximity and spectral similarity between units. After clustering, the system evaluates the internal connectivity of each formed spatial cluster, specifically by calculating the proportion of core units within the cluster or the tightness of connections between units. This calculated value is compared with a second preset threshold. If it is higher than this threshold, the cluster is determined to be a valid spatial cluster with strong internal connectivity, and its corresponding candidate change group is retained; otherwise, it is discarded. Through this process, a preliminary set of selected change groups is formed.

[0182] As can be seen, through the aforementioned spatial clustering analysis and connectivity threshold judgment, the system has verified the spatial entity nature of the candidate groups initially identified based on temporal similarity. This method can effectively identify those changing regions that truly possess close adjacency relationships and internal structural coherence in spatial distribution. This ensures that the groups ultimately entering subsequent analysis not only have similarities in temporal evolution patterns but also possess the spatial basis for management as unified geographical entities, avoiding the misclassification of spatially discrete and fragmented unit sets as homogeneous regions and improving the practical operability of the identified changing regions in land spatial planning applications.

[0183] The determination module 205 performs time-series consistency analysis on each group in the preliminary filtered change group set, and integrates its interannual change characteristics to generate a spatiotemporal comprehensive homogeneity score. If the score is higher than the third preset threshold, the group is determined to be a homogeneous change region.

[0184] Specifically, the system extracts the spectral reflectance time-series curves of all spatial monitoring units within each group in the initially screened set of change groups. Using dynamic time-warped distance or K-means algorithms as grouping algorithms, the system performs secondary clustering analysis on the time series shapes within each group, identifying different change rhythms or patterns, thus obtaining multiple time-series sub-patterns and their representative units. Subsequently, the system calculates the silhouette coefficient of the entire group based on these representative units. This coefficient quantifies temporal consistency by comparing the similarity of each unit with other units within the group and its difference from the nearest external group. Simultaneously, the system calculates the interannual gradient evolution slope value of the annual seasonal fluctuation series of spectral indicators within the group; this slope reflects the strength of the change trend. Finally, the system integrates the silhouette coefficient and interannual gradient slope value into a single spatiotemporal comprehensive homogeneity score using a weighted fusion algorithm.

[0185] As can be seen, by integrating the aforementioned temporal consistency analysis with the quantitative fusion of interannual variation characteristics, the system achieves an objective and comprehensive assessment of the homogeneity within change groups. This method not only identifies the inherent consistency of groups in their temporal evolution rhythm but also considers the intensity characteristics of change trends, thus avoiding the biases that may arise from single-dimensional assessments. The introduction of a spatiotemporal comprehensive homogeneity score provides a unified and quantifiable standard for determining homogeneous change regions, ensuring that the ultimately identified regions possess a high degree of inherent consistency in both temporal and spatial dimensions, providing a reliable foundation for subsequent evolutionary pattern analysis and precise control.

[0186] The identification module 206 identifies the dominant change pattern sequence of each homogeneous change region and classifies the evolution pattern type of the region based on the regularity of the sequence.

[0187] Specifically, for each identified homogeneous change region, the system analyzes multiple temporal sub-pattern representative units obtained through time series shape clustering. Based on the order of appearance of these representative units on the time axis and their transformation relationships, the system constructs a pattern sequence that reflects the dominant change process of the region. Based on this sequence, the system extracts key change direction features and uses time series pattern recognition algorithms to analyze the repetition frequency and periodicity intensity of the main change directions in the sequence. Specifically, the repetition frequency is calculated by counting the number of times a specific pattern appears within the observation time window, and the periodic stability of the change pattern is evaluated using a periodic detection algorithm. Finally, based on preset classification rules, such as classifying regions with high repetition frequency and strong periodic stability as cyclical and regions where changes occur concentrated in short periods as sudden, the system completes the automated classification and identification of the evolutionary pattern type of each homogeneous change region.

[0188] As can be seen, through the in-depth mining and regular quantitative analysis of the dominant change pattern sequence, the system has achieved a precise interpretation of the internal evolutionary logic within homogeneous change regions. This method can not only identify the static attribute of "homogeneity" of a region, but also further diagnose the core patterns and regularities of its dynamic evolution, such as whether it exhibits periodic fluctuations or unidirectional gradual development. This provides crucial insights into understanding the intrinsic driving forces of regional change and predicting its future development trends, enabling land change analysis to move from static description of the current situation to dynamic interpretation of mechanisms, ultimately supporting the formulation of precise and differentiated spatial management strategies based on differences in evolutionary patterns.

[0189] The grouping optimization module 207 optimizes the grouping of all homogeneous change regions based on the evolutionary pattern type to obtain the final change region grouping result.

[0190] Specifically, the system acquires the classification results of evolutionary patterns across all homogeneous change regions. It employs a grouping algorithm, such as agglomerative hierarchical clustering, to perform cluster analysis on homogeneous change regions using evolutionary pattern types as feature vectors. During clustering, the system constructs a tree-like structure and calculates the similarity between branches, specifically quantifying the degree of matching of evolutionary patterns between branches using cosine similarity or the Jaccard index. The system compares the calculated similarity values ​​with a preset threshold. If the branch similarity exceeds the threshold, all homogeneous change regions contained in that branch are merged into the same group. By iteratively executing this merging process, the system ultimately generates optimized grouping results for the change regions.

[0191] As can be seen, through the above-described grouping optimization process based on evolutionary patterns, the system achieves refined and scientific merging of homogeneous change regions. This method ensures that the final grouping results not only reflect the similarity of regions in spatiotemporal characteristics but also fully consider the consistency of their inherent evolutionary dynamics. This makes the grouping results more logically sound and stable, enabling them to more accurately reveal the overall pattern and internal connections of regional changes, and providing a reliable and intuitive basis for regional division for subsequent differentiated management and planning decisions regarding national land space.

[0192] Example 3

[0193] Please see Figure 3 , Figure 3 This is a schematic diagram of another land change analysis system based on spatiotemporal data disclosed in an embodiment of the present invention. Figure 3 As shown, the device may include:

[0194] Memory 301 storing executable program code;

[0195] Processor 302 coupled to memory 301;

[0196] The processor 302 calls the executable program code stored in the memory 301 to execute some or all of the steps in the land change degree analysis method based on spatiotemporal data disclosed in Embodiment 1 of the present invention.

[0197] Example 4

[0198] This invention discloses a computer storage medium storing computer instructions. When these computer instructions are invoked, they are used to execute some or all of the steps in the land change degree analysis method based on spatiotemporal data disclosed in Embodiment 1 of this invention.

[0199] Example 5

[0200] This invention discloses a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause a computer to perform the steps of a land change degree analysis method based on spatiotemporal data as described in Embodiment 1.

[0201] The system embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0202] Through the detailed description of the above embodiments, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0203] Finally, it should be noted that the land change analysis method and system based on spatiotemporal data disclosed in the embodiments of the present invention are merely preferred embodiments of the present invention and are only used to illustrate the technical solutions of the present invention, not to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for analyzing the degree of land change based on spatiotemporal data, characterized in that, The method includes the following steps: S1. Acquire multi-temporal remote sensing data from multiple spatial monitoring units and extract the temporal characteristics of each unit; S2. Based on the temporal characteristics, analyze the temporal similarity between the spatial monitoring units, and select unit pairs with temporal similarity higher than a first preset threshold to form a preliminary set of temporal similarity pairs; S3. Based on the preliminary time similarity pair set, construct a connection relationship diagram between spatial monitoring units, and identify connected regions with a number of units greater than a preset scale threshold from the diagram as candidate change groups; S4. Perform spatial clustering analysis on the candidate change groups, and filter out groups whose internal spatial connectivity is higher than the second preset threshold to form a set of change groups after preliminary screening. S5. Perform time-series consistency analysis on each group in the preliminary set of change groups, and integrate their interannual variation characteristics to generate a spatiotemporal comprehensive homogeneity score; if the score is higher than the third preset threshold, the group is determined to be a homogeneous change region. S6. Identify the dominant change pattern sequence for each homogeneous change region, and classify the evolution pattern type of the region based on the regularity of the sequence; S7. Based on the aforementioned evolutionary pattern type, all homogeneous change regions are grouped and optimized to obtain the final change region grouping results.

2. The method for analyzing the degree of land change based on spatiotemporal data according to claim 1, characterized in that, The steps in S1 include: The study area is divided into regular grids using a spatial grid division method, with each grid serving as a spatial monitoring unit. Acquire multi-temporal remote sensing images of each space monitoring unit, and perform geometric registration, radiometric calibration and atmospheric correction to generate a surface reflectance image sequence; From the surface reflectance image sequence, pixel-level spectral reflectance time-series curves are extracted, and at least one spectral index among the normalized vegetation index and water index is calculated, thereby generating an annual seasonal fluctuation sequence of spectral indexes. The pixel-level spectral reflectance time series curve is combined with the annual seasonal fluctuation sequence of the spectral index to obtain a complete time series feature set at the unit level.

3. The method for analyzing the degree of land change based on spatiotemporal data according to claim 1, characterized in that, Step S2 includes: Calculate the Euclidean distance of temporal shape similarity between each pair of spatial monitoring units and generate a temporal similarity distance matrix; The unit pairs whose Euclidean distance is lower than the fourth preset threshold are marked as preliminary time similarity pairs to form a preliminary set; Extract the key inflection points of the time series of unit pairs in the preliminary set, and use the dynamic time warping algorithm to calculate the matching distance between the inflection points; Unit pairs whose matching distance is lower than the fifth preset threshold are retained to form the final time similarity pair set, which serves as the initial time similarity pair set.

4. The method for analyzing the degree of land change based on spatiotemporal data according to claim 1, characterized in that, The steps in S3 include: A spatial connection graph is constructed using spatial monitoring units as nodes and unit pairs in the preliminary time similarity pair set as edges; All connected components are extracted from the graph using a depth-first search algorithm; Connected components containing more than a preset size threshold are identified as candidate variation groups, wherein the preset size threshold is dynamically set based on the ratio of the total area of ​​the study area to the average area of ​​the units.

5. The method for analyzing the degree of land change based on spatiotemporal data according to claim 1, characterized in that, The steps in S4 include: Obtain the annual seasonal fluctuation series of the boundary coordinates and spectral indices of the internal units of the candidate change groups; A density-based spatial clustering algorithm is used to perform spatial neighborhood expansion clustering on the units within the group, where the neighborhood radius is adaptively adjusted according to the average geographical distance of the units. If the spatial clusters formed by clustering have internal connectivity higher than the second preset threshold, the candidate change group is retained; otherwise, it is discarded, resulting in a set of change groups after preliminary screening.

6. The method for analyzing the degree of land change based on spatiotemporal data according to claim 1, characterized in that, The steps in S5 include: Extract the time-series spectral reflectance curves of all units within the variation group, and perform time-series shape clustering using dynamic time warping distance or K-means algorithm to obtain multiple time-series sub-patterns and their representative units; Based on the representative unit of the time-series sub-pattern, the overall profile coefficient of the group is calculated as a time-series consistency index; Calculate the slope value of the interannual gradient evolution characteristics of the annual seasonal fluctuation series of spectral indicators within the group; The contour coefficients and slope values ​​are weighted and fused to generate a group spatiotemporal comprehensive homogeneity score; Groups with scores higher than a third preset threshold are identified as homogeneous change regions.

7. The method for analyzing the degree of land change based on spatiotemporal data according to claim 1, characterized in that, Step S6 includes: Based on the order of occurrence and transformation relationship of representative units of time-series sub-modes within homogeneous change regions, an annual summary of dominant change pattern sequences is generated. Analyze the repetition frequency and periodic stability variance of the main changing directions in the sequence summary; Based on the rule that the repetition frequency is higher than the sixth preset threshold and the periodic stability variance is lower than the preset variance threshold, the evolutionary patterns are classified into sudden, gradual, cyclic, or mixed types.

8. The method for analyzing the degree of land change based on spatiotemporal data according to claim 1, characterized in that, The steps in S7 include: A cohesive hierarchical clustering algorithm is used to cluster homogeneous change regions based on the type of evolutionary pattern. Calculate the cosine similarity or Jaccard index between branches of the tree structure during clustering; The regions contained in branches with a similarity higher than the fourth preset threshold are merged to obtain the optimized grouping result.

9. A land change analysis system based on spatiotemporal data, used to implement the land change analysis method based on spatiotemporal data as described in any one of claims 1-8, characterized in that, The system includes: The data acquisition module is used to acquire multi-temporal remote sensing data from multiple spatial monitoring units and extract the temporal characteristics of each unit. The analysis module is used to analyze the temporal similarity between the spatial monitoring units based on the temporal characteristics, and to filter out unit pairs with temporal similarity higher than a first preset threshold to form a preliminary set of temporally similar pairs. The generation module is used to construct a connection graph between spatial monitoring units based on the preliminary time similarity pair set, and to identify connected regions with a number of units greater than a preset scale threshold from the graph as candidate change groups; The filtering module is used to perform spatial clustering analysis on the candidate change groups, filter out groups whose internal spatial connectivity is higher than a second preset threshold, and form a set of change groups after preliminary filtering. The determination module is used to perform temporal consistency analysis on each group in the initially screened set of change groups, and integrate their interannual variation characteristics to generate a spatiotemporal comprehensive homogeneity score; if the score is higher than a third preset threshold, the group is determined to be a homogeneous change region. The identification module is used to identify the dominant change pattern sequence of each homogeneous change region and classify the evolution pattern type of the region based on the regularity of the sequence; The grouping optimization module is used to optimize the grouping of all homogeneous change regions based on the evolutionary pattern type, and obtain the final change region grouping result.

10. A land change analysis system based on spatiotemporal data, characterized in that, The system includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute a land change analysis method based on spatiotemporal data as described in any one of claims 1-8.

Images