Traditional village cluster landscape feature similarity construction system and method

Abstract

The invention discloses a traditional village cluster landscape feature similarity construction system and method, and relates to a traditional village landscape similarity construction and network analysis method based on a minimum landscape unit (MSU). According to the method, on the basis of MSU, building, ecological, cultural and other features are extracted, and village feature vectors are constructed; the comprehensive similarity between villages is calculated through five dimensions of attributes, vision, structure, pattern and time sequence, and theoretical landscape consistency is reflected. On the basis, visual, cultural and spatial adjacent constraints are introduced, actual landscape connection strength is constructed, a circuit theory is adopted to map the core value into node voltage, a landscape influence propagation process is simulated, and resistance and current strength between nodes are calculated. The model can identify ideal landscape clusters, key cultural galleries and potential breaking points, and quantitative analysis and restoration priority evaluation of traditional village landscape connection are realized. The method has the advantages of being repeatable in implementation, clear in model, visual in result and the like, can remarkably improve scientificity and objectivity of cultural heritage landscape pattern analysis and traditional village centralized contiguous protection utilization planning, and has technical innovation value and application and popularization potential.

Description

Technical Field

[0001] This invention relates to computer technology, village landscape feature similarity identification and protection technology, and in particular to a system and method for constructing similarity features of traditional village cluster landscapes. Background Technology

[0002] Traditional villages refer to rural settlements and their surrounding environment that have a long development cycle, significant resource advantages, and unique landscape features. They contain rich cultural, historical, natural, and artistic values ​​and are of great significance for scientific protection, inheritance, and development.

[0003] Traditional villages serve as iconic carriers of the historical evolution of rural living environments in my country. Currently, 8,155 rural settlements are listed in the Directory of Traditional Villages of China. These villages generally possess systemic characteristics such as diverse environmental endowments, multiple natural and cultural elements, and complex development paths. Furthermore, due to their concentrated geographical distribution, similar spatial patterns, shared cultural lineages, and interconnected ecological environments, they are often not isolated entities but exhibit a holistic characteristic of forming networks from points and connecting areas from networks, possessing the attributes of clustered protection and collaborative development. However, currently, there is no technology for identifying the similarity between villages; at most, image recognition technology is used for rough similarity identification, which is far from sufficient. This leads to difficulties in subsequent cluster management, making it difficult to delineate and effectively incorporate similar villages into a cluster for unified management. Therefore, the identification of the similarity of landscape characteristics in traditional village clusters has become a prominent research issue, especially in terms of overall similarity and local similarity. Scientific analytical methods and related technologies are urgently needed to provide scientific guidance for the concentrated delineation, protection, construction, and management of traditional villages.

[0004] This invention develops a system and method for constructing similarity of landscape features in traditional village clusters. It is a method for constructing and analyzing the similarity of landscape features in traditional villages based on the minimum landscape unit (MSU). Through a systematic modeling path of "MSU fine representation—multi-dimensional similarity fusion—circuit propagation simulation," this method achieves quantitative identification, networked expression, and visualized priority assessment of traditional village landscape features. It possesses advantages such as repeatability, clear models, and visualized results, significantly improving the scientific rigor and objectivity of cultural heritage landscape pattern analysis and the planning of concentrated and contiguous protection and utilization of traditional villages. It has technological innovation value and application potential. Summary of the Invention

[0005] In view of the above-mentioned deficiencies of the prior art, the technical problem to be solved by the present invention is to provide a traditional village cluster landscape feature phase.

[0006] A similarity construction system and method that can accurately identify the similarities between villages.

[0007] To achieve the above objectives, the present invention provides a system for constructing similarity of landscape features of traditional village clusters, comprising:

[0008] The multi-source data acquisition and preprocessing module, as the system's data entry point, is responsible for acquiring and standardizing various types of raw data, and cleaning, removing, and organizing the data.

[0009] The automatic minimum landscape unit segmentation module divides a continuous study area into the smallest, landscape-significant minimum landscape units (MSUs) in a physically or visually separable manner.

[0010] The landscape feature vector construction module is responsible for aggregating discrete MSU information into a quantitative description representing the overall attributes of the village. First, MSUs are assigned to villages based on their spatial location. Then, multidimensional features of each MSU are extracted. An aggregation algorithm is used to aggregate the MSU features to the village scale, standardization is performed, and the features are assembled in a fixed order to form the characteristics of each village. Standardized landscape feature vector , It is the core data for subsequent macro analysis and is simultaneously supplied to the landscape core value calculation module and the multidimensional similarity calculation module.

[0011] The landscape core value calculation module is used to quantify the potential and importance of villages within a region; the input landscape feature vector construction module provides... ,from Extracting specific indicators ( The core value of each village is calculated using a geometric mean model and then output. ; These are the key driving force data, which are supplied to the circuit theory simulation module. Transformation of the "voltage source" in ideal landscape network simulation ”;

[0012] The multidimensional similarity calculation module is responsible for calculating the comprehensive similarity between villages in terms of multidimensional landscape features. By parallel computing five different similarity components, a comprehensive similarity assessment is constructed: ontological attribute similarity, visual image similarity, topological / structural similarity, spatial pattern similarity, and temporal similarity. Subsequently, each similarity component is integrated into a comprehensive similarity to reflect the overall similarity between villages.

[0013] This invention also discloses a method for constructing similarity of landscape features of traditional village clusters, comprising the following steps:

[0014] S100, Data Acquisition, Processing and Segmentation MSU

[0015] Multi-source data, including elevation data, image data, vector base maps, and socio-cultural data, were collected and preprocessed and standardized. MSUs were delineated automatically and with high precision using a progressive strategy. The study area was divided into the smallest landscape units through an automated delineation strategy, which served as the basic spatial units for landscape feature analysis.

[0016] S200, Multidimensional Feature Extraction

[0017] Transforming the multidimensional and heterogeneous landscape information of villages into unified and standardized numerical vectors. The system extracts multidimensional features of each MSU to achieve digital representation and quantitative comparison of landscape features, providing a concrete and operable data foundation for subsequent core value calculation, similarity analysis, network construction and spatial simulation.

[0018] S300, Landscape Feature Vector Construction

[0019] Based on the S130 high-precision minimum landscape unit (MSU) division, the multidimensional and heterogeneous landscape information of villages is transformed into a unified and standardized numerical vector. ;

[0020] S400, Calculation of Core Value of Villages

[0021] Based on the landscape feature vector The geometric mean model is used to calculate the core landscape value of each village based on specific indicators. , Transformation of the "voltage source" in ideal landscape network simulation ”;

[0022] S500, Similarity Calculation Between Villages

[0023] Calculate ontology similarity between villages Visual image similarity ( Topological / structural similarity Spatial pattern similarity Temporal similarity ( The components are then combined to form the final overall similarity.

[0024] The beneficial effects of this invention are:

[0025] This invention proposes a method for constructing similarity and network connections of traditional village landscapes based on the minimum landscape unit (MSU). This method can achieve unified modeling between theoretical similarity and real-world spatial constraints, and has clear hierarchical logic and significant technical advantages.

[0026] First, this invention introduces the Minimum Landscape Unit (MSU) as the basic unit of analysis. Compared with the traditional approach that uses administrative boundaries or village areas as the analysis scale, this method can more precisely characterize the spatial heterogeneity of landscape elements within villages, ensuring that the feature vectors have spatial resolution and repeatability. Through area-weighted aggregation and standardization, a structured feature matrix that can be compared across villages is formed, providing a unified input for subsequent similarity modeling.

[0027] Secondly, this invention constructs a multidimensional similarity model to achieve full-dimensional similarity evaluation from attributes, visual aspects, structure, pattern, to temporal sequence. Each component is independently modeled using algorithms such as entropy weighting, spectral analysis, Earth Distance Moving (EMD), and Dynamic Time Warping (DTW), and a comprehensive similarity matrix is ​​obtained through weight fusion. It can identify potential landscape homogeneity among villages in different dimensions.

[0028] Furthermore, this invention innovatively introduces a spatial propagation model based on circuit theory, mapping the core value of the landscape to nodal voltages, constructing geographical and ecological factors as landscape resistance surfaces, calculating landscape currents using Ohm's law, and building a simulation network of landscape accessibility and energy flow. This model can assess the spatial connectivity and propagation potential of traditional village systems from a holistic perspective, breaking through the limitations of previous analyses that could only be performed locally.

[0029] Furthermore, this invention establishes a current landscape connection strength model, introducing three types of real-world constraints—visual, cultural, and spatial proximity—on the basis of theoretical similarity, thereby realizing the mapping from "potential similarity" to "real-world connection strength". This represents the theoretical level of multidimensional landscape consistency, while Reflecting the actual spatial accessibility under geographical accessibility constraints, the comparison between the two can directly identify "potential breakpoints" and "potential cultural corridors," providing a quantitative basis for landscape restoration.

[0030] In summary, this invention, through a systematic modeling path of "MSU fine representation—multidimensional similarity fusion—circuit propagation simulation," achieves quantitative identification, networked representation, and visual priority assessment of traditional village landscape features. This method is repeatable, has a clear computational path, and controllable parameters, significantly improving the scientific rigor and objectivity of landscape pattern analysis and cultural heritage protection planning. Attached Figure Description

[0031] Figure 1 This is a schematic diagram of the system for constructing similarity in the landscape features of traditional village clusters;

[0032] Figures 2-11 This is a schematic diagram of the construction of resistance surfaces and identification of potential cultural heritage corridors based on circuit theory in one embodiment of the present invention; wherein, Figure 2 This is an elevation resistance surface. Figure 3 For slope resistance surface, Figure 4 For undulation resistance surface, Figure 5 For land use resistance surface, Figure 6 As a buffer resistance surface for railways, Figure 7 As a buffer surface for highway resistance, Figure 8 Serving as a buffer zone for the city's main roads. Figure 9 Serving as a buffer resistance surface for urban secondary arterial roads. Figure 10 For the resistance surface of village agglomeration; Figure 11 This is to obtain a comprehensive resistance surface by standardizing and weighting the above single-factor resistance surfaces according to preset weights; Figure 12 This is a schematic diagram illustrating the identification results of potential cultural heritage corridors calculated based on circuit theory, driven by the nodal voltages corresponding to the comprehensive resistance surface and the core value of the village. The basic data comes from traditional villages in Xindu District. Detailed Implementation

[0033] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0034] See Figure 1 A system for constructing similarity in landscape features of traditional village clusters, including:

[0035] The multi-source data acquisition and preprocessing module, as the system's data entry point, is responsible for acquiring and standardizing various types of raw data, and cleaning, filtering, and organizing the data.

[0036] The automatic minimum landscape unit (MSU) segmentation module divides the continuous study area into the smallest, landscape-significant MSUs using physically or visually separable methods. Fine-grained landscape patches are generated using imagery and elevation data provided by the multi-source data acquisition and preprocessing module, employing a progressive strategy of "two-dimensional initial segmentation, three-dimensional refinement, and model correction."

[0037] The landscape feature vector (MSU) construction module is responsible for aggregating discrete MSU information into a quantitative description representing the overall attributes of a village. First, MSUs are assigned to villages based on their spatial location. Then, multi-dimensional features (e.g., 18-dimensional / 37-dimensional, including architecture, ecology, culture, connectivity, etc.) of each MSU are extracted. These features are then aggregated to the village scale using aggregation algorithms (e.g., mean, density, proportion, diversity index, etc.), and standardized (e.g., range standardization, one-hot encoding) is performed. Finally, the features are assembled in a fixed order to form the characteristics of each village. Standardized landscape feature vector . It is the core data for subsequent macro-analysis and is simultaneously supplied to the landscape core value calculation module and the multi-dimensional similarity calculation module.

[0038] The landscape core value calculation module is used to quantify the potential and importance of villages within a region. The input landscape feature vector construction module provides... ,from Extracting specific indicators ( The core value of each village is calculated using a geometric mean model and then output. . These are the key driving force data, which are supplied to the circuit theory simulation module. Transformation of the "voltage source" in ideal landscape network simulation ".

[0039] The multidimensional similarity calculation module is responsible for calculating the comprehensive similarity between villages in terms of multidimensional landscape features. It does not rely on actual geographical obstacles, but rather measures the potential consistency of villages from the perspective of "feature meaning." By computing five different similarity components in parallel, a comprehensive similarity assessment is constructed: ontological attribute similarity, visual image similarity, topological / structural similarity, spatial pattern similarity, and temporal similarity. Subsequently, these similarity components are merged into a comprehensive similarity, reflecting the overall similarity between villages.

[0040] The existing landscape connection network construction module constructs an actual landscape connection strength network. That is, an objective description of "what it is like now". The input landscape feature vector construction module provides... fusion weight ,calculate (ontology similarity) (Corridor connection) (Common characteristics of the overall structure) will eventually be integrated to achieve the following. Current status and network connections (manifested as) The similarity matrix or network graph is then fed to the breakpoint and potential corridor identification module (as the "current state" side for comparison).

[0041] The circuit theory simulation module is used to simulate the ideal landscape connection pattern, i.e., "how the connections should ideally be." The input is provided by the landscape core value calculation module. (Converted to voltage), resistance surface weight (used to construct the overall resistance value) ) Calculation of ideal current using a theoretical model of the operating circuit. and ideal cluster (via Louvain algorithm), key corridors Ideal connection pattern ( ) and ideal connection strength ( The subsequent input breakpoint and potential corridor identification module serve as the "ideal" side for comparison.

[0042] The fault point and potential corridor identification module identifies problems and opportunities by comparing the "current state" with the "ideal state," and provides decision-making suggestions. By comparing the actual landscape connection strength network constructed by the current landscape connection network construction module with the ideal landscape cluster generated by the circuit theory simulation module, it identifies fault points and potential corridors in the current landscape pattern, providing a basis for formulating landscape protection and restoration strategies.

[0043] The feedback optimization module is used for the system's closed-loop optimization mechanism to achieve continuous learning and evolution.

[0044] The method for constructing the similarity of traditional village cluster landscape features in this embodiment includes the following steps:

[0045] S100, Data Acquisition, Processing and Segmentation MSU

[0046] Multi-source data, including elevation data, image data, vector base maps, and socio-cultural data, were collected and preprocessed and standardized. MSUs were automatically and accurately delineated using a progressive strategy. The study area was divided into the smallest landscape units through an automated delineation strategy, which served as the basic spatial units for landscape feature analysis.

[0047] S200, Multidimensional Feature Extraction

[0048] Transforming the multidimensional and heterogeneous landscape information of villages into unified and standardized numerical vectors. This process extracts multidimensional features from each MSU (Multi-Suite Unit) to achieve digital representation and quantitative comparison of landscape features, providing a concrete and operable data foundation for subsequent core value calculation, similarity analysis, network construction, and spatial simulation. This step is a crucial bottom-up fusion point.

[0049] S300, Landscape Feature Vector Construction

[0050] Based on the S130 high-precision minimum landscape unit (MSU) division, the multidimensional and heterogeneous landscape information of villages is transformed into a unified and standardized numerical vector. This step is crucial for achieving refined analysis: first, multidimensional features of each MSU are extracted, and then aggregation algorithms are used to aggregate the MSU features to the village scale. This not only provides a unified village-level data foundation for subsequent macro / overall analysis, but also... More importantly, it provides an indispensable microscopic spatial basis for calculating the similarity between S520 visual images and S540 spatial patterns, thereby enabling a comprehensive analysis from microscopic components to macroscopic features.

[0051] S400, Calculation of Core Value of Villages

[0052] Based on the landscape feature vector The geometric mean model is used to calculate the core landscape value of each village based on specific indicators. , Transformation of the "voltage source" in ideal landscape network simulation ".

[0053] S500, Similarity Calculation Between Villages

[0054] Calculate ontology similarity between villages Visual image similarity ( Topological / structural similarity Spatial pattern similarity Temporal similarity ( The components are then combined to form the final overall similarity.

[0055] S600, Constructing a network of actual landscape connectivity strength

[0056] By incorporating real-world constraints such as geography, culture, and ecology, a landscape connectivity intensity network reflecting the connectivity of real spaces is constructed in parallel. It is an "accessibility-weighted similarity projection", and the network focuses on actual connections under geographical accessibility constraints.

[0057] S700, Circuit Simulation and Network Propagation Analysis

[0058] Based on circuit theory, this study simulates the flow and distribution of landscape influence in traditional village networks by constructing landscape resistance surfaces and calculating landscape currents and voltages, thereby identifying ideal landscape clusters and key corridors. Circuit simulation serves as the dynamic propagation layer, generating current density.

[0059] S800, by comparing the actual landscape connectivity network constructed by S600 with the ideal landscape cluster generated by S700, identifies breakpoints and potential corridors in the current landscape pattern, providing a basis for formulating landscape protection and restoration strategies.

[0060] S900 collects user feedback through an active learning mechanism, iteratively optimizes model parameters, and achieves closed-loop updates and adaptive improvements to the system. Feedback quantification method: Adjusts weights and sample selection strategies (e.g., based on uncertainty sampling) based on user ratings of clustering results. This is existing technology, and existing feedback optimization techniques can be directly adopted.

[0061] Preferably, S100 further includes:

[0062] S110, Multi-source data acquisition, including at least:

[0063]

[0064] S120. Data preprocessing, including:

[0065] Coordinate unification: Convert to a unified coordinate system (e.g., CGCS2000UTM50N), with control point deviation ≤ 0.1m;

[0066] Point cloud filtering: classifies ground points, vegetation, and buildings, with an elevation deviation ≤ 0.3m;

[0067] Vector optimization: Smooth the building outline, remove burrs ≤0.5m, and ensure a LiDAR match ≥95%;

[0068] Data import: Store in PostGIS and create a spatial index (R-tree).

[0069] S130, Divide into Minimum Landscape Units (MSUs)

[0070] Automated and high-precision generation of basic spatial units (MSUs) for subsequent analysis (such as single houses, single plots of farmland, forest patches, etc.) provides an objective, repeatable, and refined microscopic analysis basis; including:

[0071] S131. Primary 2D partitioning based on aerial images (visual / physical segmentation)

[0072] This method rapidly identifies major planar patches using image color, texture, and boundary information. It employs edge detection (Canny / Sobel), color clustering (K-meansinLab / HSV), physical boundary line recognition (Hough transform + vector constraints), and spatial constraint clustering (SC-DBSCAN, using hybrid distance). Algorithm, post-processing merging of small units ( ).

[0073] Parameter acquisition: Weights in the mixing distance formula , , (satisfy The optimal IoU (Internal Value) can be determined through cross-validation with a small sample (e.g., randomly selected 1 km² region), using manually labeled MSU boundaries as the gold standard, and optimizing the IoU metric (e.g., IoU ≥ 0.8) and sample size (e.g., 1 km²). Weights (e.g., 0.3 / 0.4 / 0.3). Minimum area threshold. Depending on the research scale and image resolution, for a DOM of 0.02m / pixel, it is usually set to 9m². The settings are determined based on the research scale and image resolution (e.g., 10m). Output: Preliminary two-dimensional partitioning. .

[0074] S131.1 Extracting Common Nodes

[0075]

[0076] S131.2 Constructing the segment chain (boundary basis)

[0077]

[0078] S131.3 MSU (Micro-unit Generation)

[0079] 1. Input the base plane:

[0080] Building MSU: Continuous building cluster surface extracted by LiDAR (area ≥ 10m²).

[0081] Vegetation MSU: DOM-interpreted NDVI > 0.3 patches (area ≥ 20m²);

[0082] Water body MSU: Independent water body (area ≥ 5m²) extracted from water system vectors;

[0083] 2. Erasure and Segmentation: Input building MSU (LiDAR extracted, area ≥10m²), vegetation MSU (NDVI>0.3, area ≥20m²), and water body MSU (area ≥5m²), and use segmentation chain to erase and generate two-dimensional MSUs, with the dominant type accounting for ≥80% of the internal data.

[0084] 3. Attribute assignment: ID (e.g., HC-MSU-001), type (building / vegetation / water body), area, center point coordinates (CGCS2000), cultural association (whether it includes cultural heritage sites);

[0085] 4. Quality Control:

[0086] Boundary deviation: 50 MSUs were randomly measured using GNSSRTK, and the deviation was ≤1m;

[0087] Internal consistency: The dominant type accounts for ≥80% within the MSU (e.g., building area ≥80% within the architecture MSU);

[0088] Scale qualification rate: MSUs with a size of 10-1000m² account for ≥90%.

[0089] S132, Height partitioning based on LiDAR / stereo imaging (3D refinement)

[0090] By utilizing elevation information to refine zoning, hierarchical structures that cannot be distinguished by 2D imagery are addressed, improving the 3D rationality of the MSU. A 1mDSM / DTM is generated based on LiDAR, and calculations are performed. .like (e.g., 0.5m²) or (e.g., 2.0m) will trigger.

[0091] The K-means algorithm (n=2-4, 100 iterations) is effective for... Layered, DBSCAN (eps=2m,) is executed within each layer. This involves splitting the data into 3D heterogeneous units (such as "ground level - first floor of the building"). The result is a refined 3D partition. .

[0092] S133, 3D Re-identification and Error Correction (Model Correction)

[0093] The partitioning results are matched and verified against the village's 3D model to correct topological errors and boundary deviations. Method: Boundary matching ( (Adjust if 1.0m); perform topology checks (overlap rate > 5%, gap > 1m) and correct; optimize boundaries using GNN; provide an expert interactive correction interface. The final output is an MSU set. .

[0094] GNN Boundary Optimization: For MSU boundaries with a deviation > 1m, a lightweight graph neural network (GNN) is used, employing a lightweight graph convolutional network (GCN). The input is the coordinates of the MSU boundary vertices. The hidden layer dimension is 32, the number of iterations is 500, and the loss function is the sum of squared boundary deviations. The optimization stops when the loss value changes by ≤1e-6 over 10 consecutive iterations to ensure that the boundary matches the village's 3D model by ≥95%.

[0095] Expert Interaction: Provides an ArcGIS editing interface for manual fine-tuning of complex landscapes (such as MSUs around ancient bridges), with the adjustment parameters automatically recorded in the log.

[0096] Quality verification: Randomly sample 5-10% MSU and compare with the manually labeled gold standard. IoU ≥ 0.8, boundary deviation ≤ 1m.

[0097] Preferably, S200 further includes:

[0098] S210, Data Acquisition: Acquire spatial data, attribute data, and expert knowledge defined in S100.

[0099] S220, Definition and Quantification of Indicator System (5 major categories, 30 basic indicators, expanded to 37 dimensions)

[0100] S221, Architectural Form (11 Dimensions)

[0101]

[0102] S222, Spatial Texture (8 Dimensions)

[0103]

[0104] S223, Landscape Environment (6 dimensions)

[0105]

[0106] S224, History and Culture (6 dimensions)

[0107]

[0108] S224, Socioeconomic (6 dimensions)

[0109]

[0110] S300, Landscape Feature Vector Construction

[0111] S310, Feature Vector Construction

[0112] First, perform data cleaning on the original attributes of MSU (fields with a missing value > 5% are imputed using KNN or expert input; KNN interpolation uses Euclidean distance as the metric, e.g., k=5, implemented using the Python scikit-learn library), and truncate outliers (IQR 1.5). All raw values ​​must undergo missing value handling, outlier truncation, and MinMax or Z-score standardization before entering the calculation. Feature vector construction formula:

[0113]

[0114] in:

[0115] The landscape feature vector of village (i) is obtained by integrating multi-source data (remote sensing, questionnaires, statistics, LBS, literature);

[0116] The standardized value of village (i) on the (k)th feature index, each index is first normalized to [0,1] or one-hot encoded;

[0117] n, the number of features (usually 37), includes five categories of indicators: building density, roof material, road tortuosity, ecological greenness, and number of cultural heritage sites.

[0118] Missing value handling: KNN interpolation (k=5, based on MSU spatial proximity) is used when the missing value rate is <5%, and data is re-collected when the missing value rate is ≥5%; Sensitivity analysis: PCA dimensionality reduction, the cumulative variance of the top 10 principal components is ≥85%; Field validation: Data quality check; Sensitivity analysis; Field sampling validation, the consistency between the eigenvalues ​​and the field survey is >0.85.

[0119] S320, Aggregated Village Characteristics (Weighted Aggregation by MSU Area):

[0120]

[0121]

[0122] : MSU(p) area (m²), calculated from vector polygons;

[0123] : The kth attribute of MSU (normalized, range [0,1]), such as NDVI, building density, cultural point density, etc.

[0124] S400, core landscape value, landscape voltage

[0125] S410, Core Landscape Value

[0126] To quantify the potential of villages as core elements of landscape conservation, a geometric mean model was used to calculate the core landscape value of each village i. This model, originating from ecology and economics, describes the overall output of a system under the synergistic effect of multiple factors. Its core emphasis is on the balance and indispensability of each factor; that is, a severe shortage of any factor will lead to a sharp decline in overall value, perfectly aligning with the principle of "value integrity" in cultural heritage protection. The calculation formula is as follows:

[0127]

[0128] in:

[0129] A comprehensive index of architectural form / spatial texture, taking... Standardized value for B-01 (proportion of traditional buildings);

[0130] Historical and cultural indices The standardized value of C-01 (cultural relic protection level value) in China. ,( =National-level cultural relic protection × 4 + ... + County-level × 1);

[0131] Landscape environment or socio-economic impact index. The standardized value of C-02 (Intangible Cultural Heritage Index) in China. ,( =World-class intangible cultural heritage × 5 + ... + County-level × 1);

[0132] Quality control: The correlation coefficient between the data and expert scores (1-5 points) is ≥0.85 using Pearson correlation coefficients. This is based on macroscopic village feature vectors. Computation can emphasize the integrity of value and provide a driving force for dynamic simulation.

[0133] S420, Transforming the core value of the landscape into landscape voltage The calculation formula is as follows:

[0134]

[0135] , The highest and lowest core values ​​in the sample set are taken respectively.

[0136] S500, Similarity Calculation

[0137] In a unified feature space (village feature vectors aggregated from MSUs) On the surface, calculate the potential landscape similarity matrix between villages. It describes whether villages are "similar in a characteristic sense," without taking into account actual geographical resistance.

[0138] The S500 module calculates a theoretical landscape feature similarity matrix. This matrix does not consider actual geographical barriers and spatial connectivity constraints, but only reflects the potential similarity relationships of villages in a multi-dimensional feature space. The results will be further weighted in the S600 module after introducing geographical resistance and corridor constraints to generate the "current landscape connectivity strength". This allows for the mapping from theoretical similarity to real-world connections.

[0139] S510, Similarity of Village Attributes

[0140] It measures the consistency of villages across cultural, architectural, ecological, and socioeconomic attributes, providing interpretable benchmark components. These include:

[0141] S511. Entropy weight calculation (determining the initial weight for each dimension):

[0142]

[0143]

[0144] in:

[0145] The feature matrix is ​​taken from S200; all feature columns are standardized to the [0,1] interval before calculation.

[0146] The information entropy of feature (k) measures the degree of dispersion of this indicator;

[0147] The entropy weight coefficient is inversely proportional to the information entropy; the greater the dispersion (the stronger the discrimination), the higher the weight.

[0148] m represents the number of villages in the study area, specifically the actual number of villages included in the statistical sample.

[0149] The proportion of feature (k) in village (i) of the sample, expressed as a standardized value. The proportion of the total number of samples;

[0150] If a certain indicator has little difference across all villages, it has high information entropy, weak discriminative power, and low weight; if the difference is large, it means that the indicator can effectively distinguish different villages, and therefore has high weight.

[0151] S512, Dynamic Weight Correction:

[0152]

[0153]

[0154] in:

[0155] The target correlation coefficient (time-related) reflects the relevance of the indicator to the current task, such as increasing the weight of culture during the cultural protection period.

[0156] Dynamic weights are combined with entropy weights and the results of their association with the target, and then re-normalized.

[0157] The (k)th feature column;

[0158] The current stage's goal-oriented vector. For example, in a stage focusing on cultural heritage protection, It can be set as a vector of cultural heritage protection unit level values ​​(C-01) for all villages; in stages focusing on tourism development, It can be set as a tourist experience satisfaction (E-05) vector.

[0159] The Pearson correlation coefficient is used to measure the degree of linear correlation between the k-th feature and the current protected target.

[0160] S513, Weighted Mahalanobis Distance and Similarity

[0161] Weighted Mahalanobis distance:

[0162]

[0163] Similarity mapping:

[0164]

[0165] in:

[0166] Ontology similarity is expressed as the reciprocal of distance, with values ​​closer to 1 indicating greater similarity.

[0167] W, the weight matrix;

[0168] According to the aforementioned structure;

[0169] The covariance matrix is ​​calculated based on the landscape feature vectors of all village samples. To prevent matrix singularity issues caused by small samples or feature collinearity, a condition number check is performed before calculation. If the condition number > 1000, then... The diagonal matrix (i.e., degenerated into weighted Euclidean distance) is used for calculation.

[0170] Weighted distance, the weighted difference between the two villages in multidimensional space.

[0171] S520, Visual Image Similarity ( )

[0172] Visual feature weighting calculation based on minimum landscape units (MSUs). This component is used to enhance attribute similarity and reflect the consistency of villages in terms of visual appearance and perception. Unlike traditional methods that directly extract features from the entire village image, this invention achieves refined visual analysis based on MSU division, using landscape component weighting. This more accurately reflects the overall visual characteristics of villages composed of different functional / morphological patches. The calculation formula is as follows:

[0173]

[0174] in: The visual feature vector of village (i) must be calculated based on MSU partitioning. The specific process is as follows:

[0175] 1. MSU image patch feature extraction:

[0176] For each MSU polygon corresponding to an aerial or satellite image patch, extract its visual features. This includes:

[0177] Color characteristics: HSV color space three-channel histogram (32 levels per channel).

[0178] Texture features: Energy, entropy, contrast, homogeneity and other indicators are calculated based on the gray-level co-occurrence matrix (GLCM, orientation 0°, 45°, 90°, 135°, distance 1 pixel).

[0179] Morphological characteristics: compactness and shape index of the building MSU.

[0180] 2. Area-weighted pooling: The visual feature vectors of all MSUs within village i are weighted according to their area. A weighted average is performed to aggregate the visual feature vectors at the village level. :

[0181]

[0182] in It is the visual feature vector of a single MSU; It refers to the area of ​​MSU.

[0183] 3. Standardization: The final result The vector is normalized to the interval [0,1].

[0184] Visual similarity is calculated by assigning a higher value to a more consistent visual appearance. This method ensures that the visual similarity of a village is a weighted comprehensive representation of the visual characteristics of its various landscape components, and its accuracy directly depends on the accuracy of the MSU classification in S130.

[0185] S530, Topological / Structural Similarity (Measuring similarity in spatial organization)

[0186] Used to measure the spatial organization consistency of buildings and road networks within a village.

[0187] S531. Construct a village spatial map model:

[0188]

[0189] in:

[0190] : A set of nodes representing the center point of a building or the intersection of roads;

[0191] Edge set: Represents a spatial connectivity relationship between nodes (e.g., connected by roads or with building spacing less than a threshold). );

[0192] The adjacency matrix is ​​defined as:

[0193]

[0194] in:

[0195] :node The Euclidean distance between them;

[0196] Distance attenuation coefficient, taken as the average building spacing in the village or an empirical value of 30–80 m;

[0197] Building connectivity threshold, taken as 50–100m, determined empirically based on plot density.

[0198] S532, Laplacian Spectrum Feature Extraction

[0199] Based on adjacency matrix Construct the degree matrix:

[0200] Calculate the normalized Laplace matrix:

[0201]

[0202] Then, feature decomposition is performed:

[0203]

[0204] in:

[0205] : Eigenvector matrix;

[0206] :forward The Laplacian spectral vector of village i, derived from the smallest non-zero eigenvalues ​​(taking the first k eigenvalues, k is recommended to be 8–12), is obtained from the local map of the village. (Nodes are buildings / intersections, edges are road connections) Calculated; Laplacian is calculated by constructing an adjacency matrix (edge ​​weights can be road class or the reciprocal of distance);

[0207] To ensure numerical stability, Perform normalization:

[0208]

[0209] S533, Spectral Similarity Calculation

[0210] Measuring the structural consistency between two villages in spectral space:

[0211] in:

[0212] Euclidean distance between the spectral characteristics of the two villages;

[0213] : Structural similarity, an exponentially decaying form of spectral distance, between 0 and 1. When the road and building networks of a village have similar connectivity patterns, the spectral distance is small, indicating high similarity.

[0214] S540, Spatial Pattern Similarity (Measuring layout / pattern consistency)

[0215] Measuring the consistency of the overall spatial distribution of MSUs (core-satellite, dispersed-centralized, etc.), based on the Earth's travel distance (EMD) of geographical distribution:

[0216]

[0217]

[0218] in:

[0219] The MSU quality allocation corresponding to village i is composed of the ratio of the area of ​​each MSU divided by S130 to the total area of ​​the village. This allocation directly reflects the spatial quality distribution of landscape components within the village and is the basis for calculating pattern similarity.

[0220] d(x,y), the geographical distance or cost distance between MSU centers;

[0221] γ is the attenuation coefficient, which depends on the average distance of the region. The default calculation method is: ,in This represents the average distance of all villages in the sample to the EMD. When the scale of the study area is large, γ can range from 0.05 to 0.2; the larger the scale, the smaller the value.

[0222] Spatial pattern similarity reflects the consistency between patch layout and pattern. EMD can be solved using Sinkhorn or linear programming; for acceleration, approximate calculations can be performed on a 10–30m grid.

[0223] The solution employs Sinkhorn optimization to accelerate EMD calculations.

[0224] S550, Temporal similarity ( (Measures evolutionary synchronicity)

[0225] Compare the temporal evolution similarity (DTW) of the two villages on key dynamic indicators (NDVI, tourist flow, etc.):

[0226]

[0227]

[0228] in:

[0229] Representative time-series vector (can be MSU aggregation or village layer);

[0230] The scaling factor (usually 1 / average distance) is used to control the rate of exponential decay.

[0231] Temporal similarity reflects the synchronicity of village evolution. If the time series length is limited, downsampling or windowed time-delayed wave (DTW) can be used for key periods (such as spring / summer / autumn / winter).

[0232] S560. Combine the components to form the final similarity.

[0233] S561, Benchmark Correction (Preventing Extreme Value Dominance)

[0234] First, apply a deviation penalty to each component:

[0235]

[0236] The median of this component in the study area. Interquartile range; Recommended value: 0.1;

[0237] The degree to which the current similarity deviates from the regional benchmark; The difference from the sample median;

[0238] The interval penalty factor is typically set between 0.1 and 0.3 and determined through sensitivity testing.

[0239] Corrected similarity ( , , , , This ensures that the results are not affected by extreme regional samples.

[0240] S562, Mutual Information Correction

[0241] Calculate the mutual information between the components If the redundancy is too high (I>T), then the weight of the side with the higher weight will be reduced to avoid duplicate contributions. This is considered redundant.

[0242] S563, Determining the Fusion Weights

[0243] Taking into account both the target task and data characteristics, Bayesian optimization or multi-objective optimization methods are adopted:

[0244]

[0245] in: To use the current similarity matrix The modularity Q value obtained after Louvain clustering is used to measure the quality of the clustering results.

[0246] The matching score for manually labeled samples. , The weights can be set manually. To optimize cluster consistency and expert evaluation, the algorithm uses cluster quality and expert rating consistency as objective functions, automatically optimizing the weights of each dimension to achieve adaptive updates. The Louvain clustering iteration count is 500, with a convergence threshold <1e-5.

[0247] S564, Comprehensive Similarity Calculation

[0248]

[0249] Overall similarity value The closer the matrix is ​​to 1, the more consistent the two villages are in terms of multidimensional landscape features. This matrix serves as the input basis for subsequent "real-world connection modeling". For each similarity component ( , , , , The weights of each weight The values ​​can be set according to different application objectives (such as ecological protection, landscape improvement, cultural revitalization, etc.); when there is no specific objective, the weights of each component are equal by default, that is: in This represents the number of similarity components (usually 5). The weights can also be assigned by experts or adjusted empirically to reflect the relative importance of each component in the overall similarity calculation.

[0250] S570, Spatiotemporal Fusion and Uncertainty Estimation

[0251] S571, Multi-temporal Synthesis Similarity

[0252] If there is data from multiple periods, then time weighting is introduced. ):

[0253]

[0254] in,( It can be set based on the time of relevance or seasonal importance.

[0255] S572. Uncertainty quantification involves repeatedly calculating the similarity distribution using Bootstrap resampling and Monte Carlo perturbation methods to obtain confidence intervals.

[0256] Confidence level is defined as:

[0257]

[0258] S573, Similarity Comparison and Discrimination Rules

[0259] when (Usually the upper quartile is taken) and confidence level It is judged as a highly similar pair.

[0260] S600, Constructing a Landscape Connection Strength Network

[0261] Comprehensive similarity matrix Primarily used to assess the inherent similarity potential of villages in terms of characteristic attributes, it can serve as an independent analysis tool or a basis for clustering. The current landscape connectivity strength network... That is in Based on this, the focus is on introducing geospatial constraints to construct a physical network for simulating the flow of landscape influences and identifying spatial problems. This network is a necessary input for circuit simulation (S700) and breakpoint identification (S800).

[0262] This network is a concrete representation of theoretical similarity in the real world. S500 is an optional, parallel analytical path used to assess "theoretical similarity"; while S600 is a core step in constructing the "real-world connection network" necessary for subsequent analysis. Both reflect village relationships from different perspectives; S600... It is a mandatory input for subsequent steps such as circuit simulation.

[0263] This step aims to apply the "theoretical similarity" calculated by S500 to the multidimensional feature space. Projected onto the physical space constrained by actual geographical, ecological, and cultural corridors, this forms a "network of current landscape connectivity intensity." . The focus is on "the strength of connections under spatial accessibility constraints," while It focuses on the "intrinsic similarity of feature attributes".

[0264] Specifically, It is composed of three core components:

[0265] Similarity of village attributes The result is directly inherited from S510 and represents the degree of similarity of the villages' intrinsic attributes.

[0266] Corridor connectivity This component is the concretization of S520 (visual image similarity) and S530 (topological / structural similarity) on real-world geographical corridors (such as sightlines and cultural routes). It quantifies the actual degree of connection between villages through existing or potential linear spatial elements.

[0267] Commonalities in Regional Patterns This component is an extension of S540 (spatial pattern similarity) onto the macro-geomorphic and ecological background of the area surrounding the village. It quantifies the consistency of the local areas where the two villages are located in terms of macro-characteristics such as topography and ecology.

[0268] S610, Calculate the connection strength of corridors

[0269] The final result of quantifying the closeness of linear spatial connections between two villages, such as visual and cultural connections, is the village corridor connection strength. This can be achieved by first calculating the visual connection strength, cultural route correlation, and spatial proximity separately, then standardizing these three types of connection strength and fusing them using a weighted geometric average. The comprehensive formula for calculating corridor connection strength is as follows:

[0270]

[0271] In the formula, The intensity of visual connection is represented and calculated using visual field analysis techniques. The degree of cultural route correlation is obtained through spatial overlay analysis; The spatial proximity is represented by a cost-distance algorithm. The weight coefficients representing the strength of connections in each dimension, and satisfying the following conditions: This reflects the importance of three corridor dimensions (visual, cultural, and spatial proximity) in the overall corridor similarity, achieving weighted fusion. The Analytic Hierarchy Process (AHP) is employed: experts in landscape planning and geographic information systems are invited to conduct pairwise comparisons of the importance of the three dimensions, constructing a judgment matrix, calculating weights, and passing a consistency test (CR < 0.1). If experts consider visual and cultural routes to be of equal importance, with spatial proximity being slightly less important, then... , , .

[0272] S611, Visual Connection Strength This method quantifies the visual connectivity of high-value MSUs (Core Landscape Units) between two villages, reflecting the degree of landscape association at the visual level. Acquisition method:

[0273] Screening high-value MSUs: Calculating the local evaluation value for each MSU In one embodiment, for building-type MSUs, The arithmetic mean of the three normalized attribute values ​​in S221—'architectural age structure,' 'building material composition,' and 'cultural association'—is used; for non-architectural MSUs, their ecological importance value is used. This is the selection process. MSU as the core landscape unit;

[0274] Viewpoint Analysis: Using ArcGIS's "Viewpoint" tool, with the observation height set to 5m (simulating the height of a human eye or landscape node), the viewpoint range (polygon) of high-value MSUs was generated.

[0275] Overlap calculation:

[0276]

[0277] in It is a village The field of view of high-value MSUs is calculated by spatial overlay analysis to determine the area ratio of intersection and union.

[0278] Both S520 (Visual Image Similarity) and S611 (Visual Connection Strength) involve visual features, but the former is based on MSU image patch aggregation, while the latter is based on visual field analysis, and their purposes are different (similarity vs. connection strength).

[0279] S612, Cultural Route Relevance This is used to quantify the spatial connectivity of cultural resources between two villages, reflecting the degree to which cultural routes support landscape connections. Acquisition method:

[0280] Cultural route data collection: Vector data of cultural routes (such as ancient post roads and intangible cultural heritage transmission routes) in the study area are obtained through literature, general surveys by cultural and tourism bureaus or field research;

[0281] Spatial Overlay: Statistical Cultural Routes High-value MSU that can be traveled through time ( )quantity ;

[0282] Density calculation:

[0283]

[0284] in It is the length of the cultural route (unit: km), which ultimately yields the density index of "number of high-value MSUs / route length".

[0285] S613, Spatial Proximity This quantifies the actual spatial accessibility of two villages, considering proximity factors such as topography and land use. Acquisition method:

[0286] Constructing dynamic resistance surfaces: Generating resistance grids by combining factors such as terrain slope, land use type, and traffic density. (The values ​​can be refined through MSU attributes, such as adjusting the building MSU resistance coefficient. Its construction principle is consistent with the landscape resistance surface described in S711. In practice, S613 (used for calculating spatial proximity) and S711 (used for circuit simulation) can share the same resistance surface, or the resistance factor weights can be fine-tuned according to specific analysis objectives (such as a greater emphasis on ecological connectivity or transportation accessibility) to generate resistance surfaces for different purposes.)

[0287] Cost Distance Calculation: Use ArcGIS's "Cost Distance" tool to calculate village distances. arrive minimum cost path resistance ;

[0288] Proximity transformation:

[0289]

[0290] Cost distance is converted into similarity (the smaller the resistance, the closer the proximity is to 1).

[0291] S620, Commonalities in Regional Patterns

[0292] The final result of quantifying the degree of consistency in macroscopic features such as topography and ecological patterns among the local areas of two villages is the commonality of village regional patterns. This can be achieved by first extracting spatial pattern-related dimensions from feature vectors, then performing feature enhancement through principal component analysis, and finally calculating cosine similarity based on the principal component scores. The formula for calculating regional pattern commonality is as follows:

[0293]

[0294] In the formula, F ik and F jk represents the projection scores of the feature vectors of villages i and j onto the k-th principal component of the spatial pattern, respectively, and p represents the number of principal components selected. S540 refers to the internal pattern of the village, and S620 refers to the regional background pattern.

[0295] S630, Calculation of Existing Landscape Connection Strength

[0296] This method projects "theoretical similarity" onto "real-world spatial connectivity." A weighted linear fusion model is used to integrate landscape relationship data from three dimensions, constructing a network of actual landscape connection strength. This network is defined as a weighted undirected graph G=(V, E, W), with each traditional village defined as a network node, forming a node set. The node set V contains all traditional villages; an edge is established between any two village nodes to form an edge set. Let E be the set of edges connecting all pairs of villages. For each edge... The weights assigned to the weighted edges are determined by the overall current connectivity strength. The formula for calculating the current landscape connectivity strength is as follows:

[0297]

[0298] In the formula, represents the strength of the existing landscape connection between villages i and j; , , Let each represent the fusion weight coefficient of each dimension, satisfying the following conditions: . , , These are the similarities in the intrinsic attributes of villages, the strength of corridor connections, and the commonalities in regional patterns. The key integration point of the unified version:

[0299] like High-potential networks with biased theoretical bias;

[0300] like High-resolution, real-world-oriented spatial constraint network;

[0301] Output That is, the weighted edge weight matrix used in circuit simulation networks.

[0302] Fusion weight coefficients of each dimension , , It can be determined using one of the following methods:

[0303] (1) Expert consultation method: Invite experts in the fields of landscape ecology and heritage protection to give independent scores and take the average value;

[0304] (2) Analytic Hierarchy Process (AHP): Construct a judgment matrix and calculate the weights;

[0305] (3) Data-driven method: Based on the initial cluster division results from experts, the optimal weights are solved in reverse through grid search or Bayesian optimization (with cluster modularity as the optimization objective). The initial recommended value is... = , =0.3, . .

[0306] S640 and Comparative diagnosis

[0307] Although the S500 module has calculated the theoretical similarity matrix between traditional villages However, this result only reflects the potential similarity relationships in the multidimensional feature space and does not consider real-world constraints such as geographical accessibility and spatial resistance. S500 represents "theoretical similarity of villages in multidimensional landscape features"; S600 represents "the strength of real-world connections under spatial accessibility constraints." High: The two villages are highly similar in terms of landscape features; Low: Although the two villages differ significantly in their characteristics; High: Strong corridors or cultural routes connecting; Low: Significantly affected by geographical or transportation barriers;

[0308] when High but When the value is low, it indicates that the two villages are similar in landscape features but are greatly affected by geographical or transportation barriers, which is a case of "potential similarity but severe barriers", and can be used as a candidate for a breakpoint.

[0309] when Low but When the value is high, it indicates that although the two villages differ greatly in characteristics, they are connected by strong corridors or cultural routes, which is a mixed corridor situation of "different types but strongly connected".

[0310] and The comprehensive judgment relationship will serve as the direct input for S810 (first round of diagnosis), providing a theoretical basis for the subsequent dual-core diagnostic model.

[0311] S700, Circuit Simulation and Network Propagation Analysis

[0312] By iteratively solving for the steady state of the entire circuit, the flow and distribution of landscape influence throughout the network are simulated, and the output is the ideal cluster partition (C) and critical corridor (KC). It calculates the global, dynamic "current density distribution".

[0313] S710, Calculate equivalent resistance

[0314] S711, Global Equivalent Resistance

[0315] A landscape resistance surface is a continuous surface that characterizes the ease with which landscape influences flow through space. In one embodiment, this invention constructs single-factor resistance surfaces based on topographic factors, land use factors, and transportation / settlement human factors, respectively, and generates a comprehensive resistance surface through standardization and weighted fusion. This comprehensive resistance surface serves as the basis for subsequent corridor identification in circuit theory. The relevant factor decomposition and results are presented in the following diagram. Figures 2-11 In one embodiment, the construction and assignment of the resistance factor specifically includes:

[0316] (1) Construction of single-factor resistance surfaces (see Figures 2-10 )

[0317] First, identify the key resistance factors affecting landscape connectivity, specifically including:

[0318] Terrain resistance factor: such as Figure 4 As shown, these include elevation resistance surfaces, slope resistance surfaces, and undulation resistance surfaces. These factors are calculated based on a digital elevation model (DEM), and corresponding resistance grids are obtained through hierarchical assignment or continuous function mapping, reflecting the physical barriers of the natural geographical environment to landscape connectivity.

[0319] Land use resistance factors: such as Figure 5 As shown, graded resistance values ​​are set according to the degree of obstruction to landscape connectivity based on land type (such as forest land, cultivated land, construction land, etc.);

[0320] Traffic network resistance factors: such as Figures 6-9 As shown, the buffer resistance surfaces are further refined into railways, highways, urban arterial roads, and urban secondary arterial roads. These various types of roads, acting as linear dividing elements, create buffer zones that impede landscape connectivity to varying degrees.

[0321] Humanistic agglomeration resistance factors: such as Figure 10 As shown, village clustering (or core density) is used to characterize human resistance (corresponding to the aforementioned cultural boundary or activity intensity). Areas with high clustering are conducive to landscape connectivity and are set as low resistance.

[0322] (2) Generation of composite resistance surface (see) Figure 11 )

[0323] The single-factor resistance surfaces mentioned above were standardized, and the weights of each factor were determined based on bibliometrics and expert consultation (or analytic hierarchy process). A comprehensive resistance surface is generated through a weighted superposition model. .like Figure 11 As shown, this combined resistance surface serves as the foundational input for subsequent circuit theory simulations. The formula for calculating the combined resistance is as follows:

[0324]

[0325] In the formula, This represents the single-path resistance of the pixel on the i-th resistance factor; The weight of the i-th resistance factor is determined by the analytic hierarchy process (AHP); n represents the total number of resistance factors.

[0326] This represents the overall resistance value for each MSU, across all local resistances. (This can be a unilateral or path-based) weighted aggregation to form an overall accessibility measure for a certain region or system. Used to assess the overall level of landscape accessibility and measure the overall connectivity of the system, the smaller the value, the stronger the landscape mobility of the village cluster. It can be used as a system performance feedback quantity for adjustment during the iterative optimization phase. By fusing weights or resistance surface parameters, adaptive convergence of the model can be achieved.

[0327] Figure 11 The darker the color (or the higher the value), the greater the landscape resistance, which is less conducive to landscape connections and cultural transmission between villages; conversely, areas with lighter colors represent low-resistance areas and are ideal bases for potential corridor crossings.

[0328] S712, Inter-node resistance

[0329] refers to villages arrive exist On the surface, the path resistance between villages is found through "cumulative cost path analysis" (such as Dijkstra's algorithm):

[0330]

[0331] Representative at Cumulative cost path algorithm running on the resistance surface.

[0332] S713, Global Connectivity Indicator

[0333] To quantitatively assess the overall connectivity level of traditional village cluster landscape networks, inter-village path resistance was calculated based on S712. Define global efficiency. As a comprehensive connectivity indicator for landscape networks, this indicator originates from graph theory and is used to measure the average efficiency of information or energy exchange between any pair of nodes in the network. The formula for calculating global efficiency is as follows:

[0334]

[0335] in:

[0336] The total number of traditional village nodes within the study area.

[0337] Village to the village The cumulative path resistance is calculated using the CostPath function in S712 on the composite resistance surface. The above calculations were performed.

[0338] Summing and traversing all unique village node pairs .

[0339] Indicator characteristics and interpretation:

[0340] The range of values ​​is .

[0341] when The closer the value is to 1, the better the overall connectivity of the landscape network, the higher the efficiency of landscape influence flow between villages, and the closer the system is to an ideal connectivity state.

[0342] when The closer the value is to 0, the more likely it is that there are a lot of high-resistance paths or broken areas in the network, making it difficult to connect villages and resulting in poor overall connectivity.

[0343] It is a key indicator for evaluating the macroscopic performance of the landscape network. It can serve as a benchmark for the initial state of the system and can also be used to quantitatively compare the effects of different conservation planning schemes (such as repairing broken points and adding potential corridors) on the overall connectivity of the network.

[0344] This indicator complements S730 (ideal landscape cluster identification) and S800 (breakpoint and potential corridor identification): the latter identifies the location of local problems, while the former assesses the overall health of the system.

[0345] S720, Calculate Landscape Current and Voltage

[0346] Each traditional village is considered a node in a circuit network, with its core landscape value represented as the node voltage. Based on circuit theory, the flow of landscape influence within the resistance network is calculated. The circuit network is solved using Circuitscape software to obtain the landscape current density and voltage distribution.

[0347] Landscape current calculations are based on a modified Ohm's law:

[0348]

[0349] In the formula, This represents the landscape current intensity from village i to village j; , These represent the landscape voltages of villages i and j, respectively, determined by the core landscape value. Obtained through standardization; The path resistance from village i to j is represented by , obtained through cumulative cost path analysis. The core value of the village landscape calculated in step S320 is integrated with three sub-factors: architecture, culture, and living heritage. It reflects the relative potential energy level of the village in the landscape network and is used to simulate the flow of landscape influence in circuit simulation.

[0350] S730, Identifying Ideal Landscape Clusters and Key Corridors

[0351] S731, Based on Landscape Current Density and Voltage Distribution

[0352] An adaptive thresholding method was used to identify ideal landscape clusters and key corridors. First, high-flow channels were identified as cultural heritage corridors based on current density distribution; second, core areas of landscape influence were identified based on voltage distribution; and finally, a community detection algorithm was used to cluster the landscape network. The formula for identifying ideal landscape clusters is as follows:

[0353]

[0354] The community partitioning algorithm based on landscape current networks employs the Louvain model to maximize the modularity Q. The adjacency matrix A... ij Build it in the following way:

[0355]

[0356] In the formula, C represents the village cluster division result; Represents the adjacency matrix of the landscape network; Indicates the intensity of the current in the landscape; This is a current intensity threshold used to filter significant landscape connections; typically, all values ​​can be selected. The upper quartiles.

[0357] S732. Identification of critical corridors is based on current density maps:

[0358]

[0359] In the formula, KC represents the set of key corridors; This represents the current density value between i and j; Indicates the current density threshold; Indicates path resistance; This indicates the resistance threshold.

[0360] S733, Corridor Identification Results (see...) Figure 12 )

[0361] based on Figure 11 The overall resistance surface is calculated using circuit theory models to determine the global current density. Figure 12 The diagram showcases identified potential and critical corridors. The highlighted connectivity network clearly connects the core nodes, primarily distributed along gently sloping valleys or existing low-resistance zones, thus avoiding high-resistance barriers. This result visually demonstrates the ideal connection paths for traditional village clusters, providing a visualized "ideal" reference for determining breakpoints in the S800 network.

[0362] S800, Identifying Breakpoints and Potential Corridors

[0363] S810, Identify the break point

[0364] Landscape connectivity breakpoints refer to village pairs within an ideal landscape cluster that share highly similar intrinsic attributes but have weak current connectivity. Based on the ideal cluster division results of S700, all village pairs belonging to the same ideal cluster are identified; the current connectivity strength of these village pairs is calculated; and breakpoints are identified using threshold analysis. These breakpoints correspond to regions with high resistance but high voltage gradients in the circuit network and are potential priority nodes for landscape restoration.

[0365] S811, First Round Diagnosis

[0366] Based on the logic of S640, identify all village pairs that are "similar in features but separated by reality".

[0367]

[0368] BP1 (candidate breakpoint set 1) consists of village pairs that should be highly similar in features, but whose actual connections are weak due to geographical barriers, inconvenient transportation, and other reasons. This is a direct diagnosis of the problem of "potential similarity failing to translate into actual connection".

[0369] S812, Second Round of Diagnosis

[0370] Based on the ideal cluster partitioning of S700, identify weak links within the cluster.

[0371]

[0372] (Candidate breakpoint set 2) consists of village pairs that belong to the same cluster in the ideal model but whose current connection strength does not meet the standard. This is a diagnosis of the "internal connectivity of the ideal cluster" problem.

[0373] In the formula, Represents the set of break points; This indicates that villages i and j are classified into the same ideal landscape cluster in S700; This represents the current landscape connectivity strength calculated by S600; This represents the threshold for the strength of connections within the cluster, calculated as the lower quartile of all current connection strengths within the ideal cluster.

[0374] S813, Comprehensive Fault Point Determination

[0375]

[0376] A village pair is included in the final remediation list if it is identified as problematic in any diagnostic model. This ensures the comprehensiveness of the diagnosis. The obtained breakpoint (BP) and potential corridor (PC) results should be verified through visualization on a GIS platform. If the consistency rate between the identification results and expert surveys is ≥85%, the model is considered to have passed verification.

[0377] S820, Identifying Potential Cultural Heritage Corridors

[0378] Potential cultural heritage corridors refer to connections between different ideal landscape clusters that possess high potential for strong corridor links but whose current integrated connections are not yet fully developed. This study screens village pairs located in different ideal clusters, assesses the corridor link strength of these village pairs, and identifies potential corridors using a dual-threshold approach. The formula for identifying potential corridors is as follows:

[0379]

[0380] In the formula, PC represents the set of potential corridors; C(i)≠C(j) means that villages i and j belong to different ideal landscape clusters; This represents the current corridor connectivity strength calculated by S600; The high-intensity threshold is represented by the upper quartile of the intensity of all corridor connections. This represents the threshold for moderate connection strength, calculated as the median of all current connection strengths.

[0381] S830, Generate a landscape restoration priority map

[0382] Based on the identification of fault points and potential corridors, a prioritization assessment model is used to determine the urgency and importance of landscape restoration. This model comprehensively considers restoration benefits and implementation costs, and determines the priority of restoration actions through a multi-indicator weighted assessment. The prioritization assessment formula is as follows:

[0383]

[0384] In the formula, The score represents the repair priority score of villages for i and j, which is standardized to the [0,1] interval and used to generate a priority heatmap;

[0385] The repair benefits are represented by a weighted average of ontological attribute similarity and potential corridor strength.

[0386] Indicates implementation costs, derived from repair resistance. and spatial distance Weighted calculation;

[0387] , , , These are the weight coefficients for each indicator, ranging from 0.2 to 0.4, which can be determined using the Analytic Hierarchy Process (AHP). An example of weight settings emphasizing cultural connectivity is as follows: α=0.4 (ontological similarity), β=0.3 (potential corridors), γ=0.2 (restoration resistance), δ=0.1 (spatial distance), which can be adjusted according to specific protection objectives.

[0388] It should be noted that, unless otherwise stated, the technical or scientific terms used in this application shall have the ordinary meaning as understood by one of ordinary skill in the art to which this application pertains.

[0389] The above description is merely a preferred embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A system for constructing similarity of landscape features of traditional village clusters, characterized in that, include: The multi-source data acquisition and preprocessing module, as the system's data entry point, is responsible for acquiring and standardizing various types of raw data, and cleaning, removing, and organizing the data. The Minimum Landscape Unit (MSU) automatic segmentation module divides a continuous study area into the smallest, landscape-significant MSUs using physically or visually separable methods. The landscape feature vector construction module is responsible for aggregating discrete MSU information into a quantitative description representing the overall attributes of the village. First, MSUs are assigned to villages based on their spatial location. Then, multidimensional features of each MSU are extracted. An aggregation algorithm is used to aggregate the MSU features to the village scale, standardization is performed, and the features are assembled in a fixed order to form the characteristics of each village. Standardized landscape feature vector , It is the core data for subsequent macro analysis and is simultaneously supplied to the landscape core value calculation module and the multidimensional similarity calculation module. The landscape core value calculation module is used to quantify the potential and importance of villages within a region; the input landscape feature vector construction module provides... ,from Extracting specific indicators ( The core value of each village is calculated using a geometric mean model and then output. ; These are the key driving force data, which are supplied to the circuit theory simulation module. Voltage source for converting ideal landscape network simulation ”; The multidimensional similarity calculation module is responsible for calculating the comprehensive similarity between villages in terms of multidimensional landscape features. By parallel computing five different similarity components, a comprehensive similarity assessment is constructed: ontological attribute similarity, visual image similarity, topological / structural similarity, spatial pattern similarity, and temporal similarity. Subsequently, each similarity component is integrated into a comprehensive similarity to reflect the overall similarity between villages.

2. The system for constructing similarity of traditional village cluster landscape features according to claim 1, characterized in that, Also includes: The existing landscape connection network construction module constructs an actual landscape connection strength network. The input landscape feature vector construction module provides fusion weight ,calculate , , Ultimately, the fusion resulted in Then it is supplied to the fracture point and potential corridor identification module; The circuit theory simulation module is used to simulate the ideal landscape connection pattern, and the input is provided by the landscape core value calculation module. Converted to voltage, a comprehensive resistance value is constructed. Calculate ideal current using a circuit theory model and ideal cluster Key corridors Ideal connection pattern and ideal connection strength, and subsequent input breakpoints and potential corridor identification modules as the "ideal" side for comparison; The fault point and potential corridor identification module identifies problems and opportunities by comparing the "current situation" with the "ideal situation" and provides decision-making suggestions. By comparing the actual landscape connection strength network constructed by the current landscape connection network construction module with the ideal landscape cluster generated by the circuit theory simulation module, it identifies fault points and potential corridors in the current landscape pattern, providing a basis for formulating landscape protection and restoration strategies.

3. A method for constructing similarity of landscape features of traditional village clusters, characterized in that, Includes the following steps: S100, Data Acquisition, Processing, and MSU Segmentation: Multi-source data, including elevation data, image data, vector base maps, and socio-cultural data, were collected and preprocessed and standardized. MSUs were delineated automatically and with high precision using a progressive strategy. The study area was divided into the smallest landscape units through an automated delineation strategy, which served as the basic spatial units for landscape feature analysis. S200, Multidimensional Feature Extraction: Transforming the multidimensional and heterogeneous landscape information of villages into unified and standardized numerical vectors. Extract multidimensional features from each MSU to achieve digital representation and quantitative comparison of landscape features; S300, Landscape Feature Vector Construction: Based on the S130 high-precision minimum landscape unit (MSU) division, the multidimensional and heterogeneous landscape information of villages is transformed into a unified and standardized numerical vector. ; S400, Calculation of Core Village Value: Based on the landscape feature vector The geometric mean model is used to calculate the core landscape value of each village based on specific indicators. , Voltage source for converting ideal landscape network simulation ”; S500, Similarity calculation between villages: Calculate ontology similarity between villages Visual image similarity ( Topological / structural similarity Spatial pattern similarity Temporal similarity ( The components are then combined to form the final overall similarity.

4. The method for constructing similarity of traditional village cluster landscape features according to claim 3, characterized in that, Also includes: S600, Constructing a network of actual landscape connectivity: Introducing real-world constraints, a landscape connection strength network reflecting the connectivity of real-world spaces is constructed in parallel. ; S700, Circuit Simulation and Network Propagation Analysis: Based on circuit theory, this study simulates the flow and distribution of landscape influence in traditional village networks by constructing landscape resistance surfaces and calculating landscape currents and voltages, thereby identifying ideal landscape clusters and key corridors. Circuit simulation serves as the dynamic propagation layer, generating current density. S800 identifies breakpoints and potential corridors in the current landscape pattern by comparing the actual landscape connection strength network constructed by S600 with the ideal landscape cluster generated by S700.

5. The method for constructing similarity of traditional village cluster landscape features according to claim 3, characterized in that, S100 also includes: S110, Multi-source data acquisition: The collected data includes at least elevation data, image data, vector base maps, and socio-cultural data; S120, Data Preprocessing: Coordinate unification: Convert to a unified coordinate system; Point cloud filtering: classifying ground points, vegetation, and buildings; Vector optimization: Smoothing the building outline; Data import: Store in PostGIS and create a spatial index; S130, MSU Division Automated and high-precision generation of the underlying spatial unit (MSU) for subsequent analysis provides an objective, repeatable, and refined microscopic analysis basis; including: S131, Primary two-dimensional partitioning based on aerial images Rapidly identify major planar patches using image color, texture, and boundary information; including: S131.1 Extract common nodes, including cultural nodes, transportation nodes, and ecological nodes; S131.2 Construct segmentation chains, including street and alley axes, water system boundaries, and building cluster boundaries; S131.3 classifies MSUs, including: (1) Input the base plane: Building MSU: Continuous building cluster surface extracted by LiDAR; Vegetation MSU: DOM-interpreted NDVI > 0.3 patches; Water Body MSU: Independent water body extracted from water system vectors; (2) Erasure segmentation: Input building MSU, vegetation MSU, and water body MSU, and use segmentation chain to erase and generate two-dimensional MSU, with the dominant type accounting for ≥80% of the internal data; (3) Attribute assignment: ID, type, area, center point coordinates, cultural association; S132, Height partitioning based on LiDAR / stereo imaging: By using elevation information to refine the zoning, we can solve the problem of hierarchical structures that cannot be distinguished by two-dimensional images and improve the three-dimensional rationality of MSU. S133, 3D Re-identification and Error Correction: The partitioning results are matched and verified with the 3D model of the village to correct topological errors and boundary deviations.

6. The method for constructing similarity of traditional village cluster landscape features according to claim 3, characterized in that, The S200 also includes: S210: Collect spatial data, attribute data, and expert knowledge defined in S100; S220, Indicator System Definition and Quantification, Characteristic Categories include Architectural Form, Spatial Texture, Landscape Environment, Historical Culture, and Socio-economic; S300 first cleans the raw attributes of the MSU. All raw values ​​need to be handled for missing values, outlier truncation, and MinMax or Z-score standardization before entering the calculation; Feature vector construction formula: ； Aggregate village features : ， 。 7. The method for constructing similarity of traditional village cluster landscape features according to claim 3, characterized in that, The S400 also includes: S410, Core Landscape Values: To quantify the potential of villages as core elements of landscape conservation, a geometric mean model was used to calculate the core landscape value of each village i. The calculation formula is as follows: ； S420, Transforming the core value of the landscape into landscape voltage The calculation formula is as follows: 。 8. The method for constructing similarity of traditional village cluster landscape features according to claim 3, characterized in that, S500 calculates the potential landscape similarity matrix between villages in a unified feature space. ,include: S510, Similarity of Village Attributes : S511. Entropy weight calculation (determining the initial weight for each dimension): ， ； S512, Dynamic Weight Correction: ， ； S513, Weighted Mahalanobis Distance and Similarity: Weighted Mahalanobis distance: ； Similarity mapping: ； S520, Visual Image Similarity ( ): The calculation is based on the visual feature weighting of the minimum landscape unit (MSU), and the formula is as follows: ， ； S530, Topological / Structural Similarity : Used to measure the spatial organization consistency of buildings and road networks within a village; S531. Construct a village spatial map model: , Matrix representation (adjacency matrix): ； S532, Laplace spectral feature extraction: Based on adjacency matrix Construct the degree matrix: Calculate the normalized Laplace matrix: ； Then, feature decomposition is performed: ； To ensure numerical stability, Perform normalization: ； S533, Spectral Similarity Calculation: Measuring the structural consistency between two villages in spectral space: S540, Spatial Pattern Similarity : To measure the overall consistency of MSU spatial distribution, Earth travel distance (EMD) based on geographic distribution is used. ； ； S550, Temporal similarity ( ): Compare the temporal evolution similarity of the two villages on key dynamic indicators: ； ； S560. Combine the components to obtain the final similarity: S561, Reference Correction: First, apply a deviation penalty to each component: ； S562, Mutual Information Correction: Calculate the mutual information between the components If the redundancy is too high (I>T), then the weight of the side with the higher weight is reduced to avoid duplicate contributions; Time is considered redundant; S563. Determining the fusion weights: Taking into account both the target task and data characteristics, Bayesian optimization or multi-objective optimization methods are adopted: ； S564. Comprehensive Similarity Calculation: 。 9. The method for constructing similarity of traditional village cluster landscape features according to claim 4, characterized in that, The S600 also includes: S610, Calculate the connection strength of corridors : The final result of quantifying the closeness of the visual and cultural linear spatial connection between two villages is the village corridor connection strength. The comprehensive formula for calculating the corridor connection strength is as follows: ； S620, Commonalities in Regional Patterns : The final result of quantifying the degree of consistency in macroscopic characteristics between the local areas where two villages are located is the commonality of the village regional pattern. The formula for calculating the commonality of the regional pattern is as follows: ； S630, Calculation of the intensity of existing landscape connections: This method projects "theoretical similarity" onto "real-world spatial connectivity." A weighted linear fusion model is used to integrate landscape relationship data from three dimensions, constructing a real-world landscape connectivity strength network G=(V, E, W). Each traditional village is defined as a network node, forming a node set. The node set V contains all traditional villages; Establish connecting edges between any two village nodes to form an edge set. The set of edges E connects all pairs of villages; for each edge... The weights assigned to the edges are determined by the overall current connectivity strength. The formula for calculating the current landscape connectivity strength is as follows: 。 10. The method for constructing similarity of traditional village cluster landscape features according to claim 9, characterized in that, The S700 and S800 also include: S700, Circuit Simulation and Network Propagation Analysis: By iteratively solving the steady state of the entire circuit, the flow and distribution of landscape influence in the entire network are simulated, and the output is the ideal cluster division (C) and key corridor (KC). S710, Calculate the equivalent drag: S711, Global Equivalent Resistance: First, key resistance factors affecting landscape connectivity were identified, including topographic factors, land use factors, transportation network factors, and human / cultural agglomeration factors. Second, values ​​were assigned to each factor based on bibliometrics and expert consultation, establishing a resistance coefficient table. Finally, a weighted overlay model was used to generate a comprehensive resistance surface. The formula for calculating the comprehensive resistance is as follows: ； S712, Inter-node resistance: ； S720, Calculate landscape current and voltage: Each traditional village is considered a node in a circuit network, its core landscape value is represented as the node voltage, and the landscape current is calculated based on a modified Ohm's law: ； S730, Identifying Ideal Landscape Clusters and Key Corridors: S731, Based on landscape current density and voltage distribution: An adaptive thresholding method was used to identify ideal landscape clusters and key corridors. First, high-flow channels were identified as cultural heritage corridors based on current density distribution. Second, core areas of landscape influence were identified based on voltage distribution. Finally, a community detection algorithm was used to cluster the landscape network. The formula for identifying ideal landscape clusters is as follows: ； The community partitioning algorithm based on landscape current networks uses the Louvain model to maximize the modularity Q, where the adjacency matrix A... ij Build it in the following way: ； S732. Identification of critical corridors is based on current density maps: ； S800, Identifying fracture points and potential corridors: S810, Identify the break point: S811, First Round of Diagnosis: Combined with S500 S600 Identify all village pairs that are "similar in characteristics but separated by reality"; ； S812, Second Round of Diagnosis: Based on the ideal cluster partitioning of S700, identify the weak links within the cluster: ； S813, Comprehensive Fault Point Determination: ； S820, Identifying Potential Cultural Heritage Corridors: Potential cultural heritage corridors refer to connections between different ideal landscape clusters that have high potential for strong corridor links but whose current integrated connections are not yet fully developed. This involves screening village pairs located in different ideal clusters, assessing the corridor link strength of these village pairs, and identifying potential corridors using a dual threshold method. The formula for identifying potential corridors is as follows: ； S830, Generate a landscape restoration priority map: Based on the identification of fault points and potential corridors, a priority assessment model is used to determine the urgency and importance of landscape restoration. Taking into account restoration benefits and implementation costs, a multi-indicator weighted assessment is used to determine the priority order of restoration actions. The priority assessment formula is as follows: 。

Images