A method for retrieving enhanced cross-period historical toponym linkages and city center shift visualizations

Abstract

The application discloses a method for retrieving enhanced cross-period historical place name linkage and city center displacement visualization, and belongs to the technical field of geographic information systems, historical place name information processing and large language model auxiliary analysis. The method comprises the following steps: historical city center point data acquisition and standardized preprocessing; based on the standardized city node data, candidate place name pairs are generated between adjacent periods or preset period combinations; retrieval enhancement evidence acquisition and structured analysis of candidate place name pairs; unified scoring and cross-period place name linkage determination; city center displacement sample construction; regionalized displacement visualization based on Voronoi unit; and result output and application. The method can realize place name matching between adjacent periods, multi-period connected components and historical city entity evolution chains, and provides continuous data support and analysis basis for long-term tracking of historical city entities, regional city system evolution analysis and administrative division change research.

Description

Technical Field

[0001] This invention belongs to the fields of geographic information systems, historical place name information processing and large language model-assisted analysis technology, specifically involving a method for enhancing the retrieval of cross-period historical place name links and visualizing the displacement of urban centers. Background Technology

[0002] The cross-period identification of historical urban entities and the analysis of urban center migration are important aspects of historical geography research. Current technologies primarily rely on place name matching and disambiguation methods for related research, and use visualization techniques such as trajectory overlay and spatiotemporal cubes to represent the spatiotemporal evolution of historical locations. While existing platforms and methods provide fundamental support for historical location labeling, association organization, and cross-period comparison, significant shortcomings remain in the automatic linking and migration trajectory visualization of cross-period urban points, including the following issues:

[0003] (1) Existing technologies focus on the construction and knowledge organization of historical place name resources, with little attention paid to the automatic linking of point-like urban data across periods, and lack mature technologies to support the continuous identification of historical urban entities.

[0004] (2) Although existing place name matching methods have gradually evolved from rule-based methods to methods that combine semantic representation and contextual information, there are still problems such as variant names, obsolete names, transliteration differences and spatial reference ambiguity in historical contexts. Relying solely on name similarity or general contextual information is still insufficient to achieve stable and reliable cross-period matching.

[0005] (3) Existing methods do not make full use of external historical documents, place name inheritance relationships and other evidence. They lack technical solutions that integrate online retrieval, structured evidence extraction, semantic scoring and spatial constraints. Therefore, in complex historical inheritance situations, the interpretability and robustness of matching results are still limited.

[0006] (4) Existing spatial visualization methods are mainly aimed at expressing individual trajectories. When the research target shifts to the identification of directional structures at the regional scale, simple trajectory overlay methods are difficult to clearly present subtle displacements and their spatial differences on small-scale maps, making it difficult to identify the directional and intensity characteristics of historical city center migration.

[0007] In summary, the names and spatial locations of historical cities across different periods change dynamically over time, and there is often no simple one-to-one correspondence between place names, administrative affiliations, and spatial locations in different periods. Furthermore, the representation of small-scale displacement objects on regional maps is easily affected by visual overlap and symbolic occlusion, and traditional visualization methods struggle to account for displacement direction, intensity, and spatial differentiation characteristics. Therefore, there is an urgent need for a cross-period historical place name linking method that integrates place name semantics, spatial constraints, and historical evidence, as well as a regionalized directional representation method suitable for small-scale displacement objects, to improve the accuracy and readability of continuous identification of historical city entities and analysis of urban center migration. Summary of the Invention

[0008] To address the shortcomings of existing technologies in automatically linking cross-period point-based city data, and the susceptibility of these technologies to factors such as variant names, obsolete names, transliteration differences, and spatial ambiguity in historical contexts, which hinder stable and continuous identification of historical city entities, and the difficulty of effectively revealing subtle displacement directions and spatial differences on small-scale maps using existing visualization methods such as trajectory overlay, this invention proposes a retrieval-enhanced method for linking cross-period historical place names and visualizing city center displacement. This method aims to solve the technical problems of insufficient reliability, weak interpretability, and unclear expression of regional migration patterns in matching results under complex historical conditions.

[0009] Technical Solution: To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0010] A method for retrieving enhanced cross-period historical place name links and visualizing urban center displacement includes the following steps:

[0011] Step 1: Acquisition and standardization preprocessing of historical city center point data;

[0012] Step 2: Based on the standardized city node data, generate candidate place name pairs between adjacent periods or preset period combinations;

[0013] Step 3: Enhanced evidence acquisition and structured analysis of candidate place name pairs;

[0014] Step 4: Standardized scoring and determination of cross-period place name links;

[0015] Step 5: Construction of urban center displacement samples;

[0016] Step 6: Visualize the regional displacement based on the Voronoi element;

[0017] Step 7: Output and Application of Results

[0018] Preferably, in step 1, the specific implementation process is as follows:

[0019] Step 101: Collect city center point data from multiple historical periods to form a cross-period city node dataset;

[0020] Step 102: Preprocess the collected city center point data and output standardized city node data.

[0021] Preferably, in step 2, the specific implementation process is as follows:

[0022] Step 201: Obtain standardized city node data for the source and target periods, and generate an initial set of node pairs to be compared using Cartesian product operation;

[0023] Step 202: Compare the place name attributes and administrative attributes of the initial node pairs to be compared. Calculate the comprehensive name score based on the similarity of standardized place names, place name core, place name pinyin, and character sequence. Calculate the administrative compatibility score by combining administrative type compatibility and administrative context similarity.

[0024] Step 203: Calculate the spherical distance between the source node and the target node based on the latitude and longitude of the node, and set an adaptive distance threshold according to the combination of node administrative category and period; combine name similarity, administrative compatibility and spatial constraints to jointly screen the initial node pairs to be compared to form a set of candidate place name pairs;

[0025] Step 204: Calculate the first-stage comprehensive score based on the comprehensive name score, administrative compatibility score, spatial score, and the hit rate of the on-site description text; for the same source node, sort the candidate place name pairs in descending order according to the first-stage comprehensive score, and output the top K candidate results to provide candidate data basis for subsequent network retrieval, evidence extraction, and cross-period place name link determination.

[0026] Preferably, in step 3, the specific implementation process is as follows:

[0027] Step 301, Enhanced Web Search: Conduct web searches from multiple sources and construct search tasks, performing search queries based on source / target location name attributes;

[0028] Step 302: Extract the structured data from the query statement using the Claude large language model;

[0029] Step 303: Score the confidence level of evidence based on the consistency of multi-source data, the authority weight of the source, and whether there is a conflict.

[0030] Preferably, in step 4, the specific implementation process is as follows:

[0031] First, the semantic matching results, external evidence credibility, and spatial consistency of candidate place name pairs are incorporated into a unified scoring framework to comprehensively evaluate and score the candidate place name pairs. The formula is:

[0032] ;

[0033] in, The semantic matching score is obtained based on the retrieval results and analysis of the large language model. The credibility score for external evidence. Spatial consistency score;

[0034] Then, the candidate place name pairs are sorted and judged according to the comprehensive score, the cross-period place name link results are determined, and multi-period connected components are further constructed to form a long-term evolution chain of historical city entities.

[0035] Preferably, in step 5, the specific implementation process is as follows:

[0036] Based on the established cross-period place name linking results, time-series connections are made between different time nodes belonging to the same historical city entity to form city center displacement samples; each displacement sample contains at least the starting point coordinates, ending point coordinates, displacement direction angle and corresponding weight information, thus providing input data for subsequent regional displacement visualization.

[0037] Preferably, in step 6, the specific implementation process is as follows:

[0038] Step 601: Aggregate the original displacement samples into Voronoi elements and record the orientation angle of each sample. and weight ;

[0039] Step 602: Calculate the dominant direction of each Voronoi element using the weighted circumferential average value;

[0040] Step 603: Map the average orientation angle to the hue component in the HSV color model to represent the migration direction;

[0041] Step 604: Normalize the average weight of samples in each unit and map it to the saturation component in the HSV color model to represent the displacement intensity.

[0042] Step 605: Finally, the color of each Voronoi cell is determined by the hue corresponding to the dominant direction and the saturation corresponding to the displacement intensity; for cells lacking effective samples or with unstable direction estimation, a no-data color is assigned.

[0043] Beneficial effects: Compared with the prior art, the present invention has the following advantages:

[0044] (1) Improve the accuracy and reliability of cross-period historical city entity identification: This invention standardizes and preprocesses the data of city center points in different historical periods, and generates candidate place name pairs by combining name similarity, spatial distance constraints and administrative compatibility, thereby enhancing the comparability between historical place name records in different periods and reducing matching errors caused by inconsistent name forms and inconsistent administrative semantics.

[0045] (2) Enhance the interpretability and robustness of matching results in complex historical contexts: This invention combines external historical evidence obtained through online retrieval with the structured analysis capabilities of a large language model, enabling cross-period place name matching to go beyond superficial comparisons of name similarity and spatial distance, and to explicitly incorporate historical semantic information such as aliases, name changes, relocation of administrative centers, and administrative translations, thereby effectively improving the reliability, interpretability, and robustness of matching decisions in complex historical contexts.

[0046] (3) Improve the comprehensive utilization of multi-source evidence: This invention constructs a unified scoring framework consisting of semantic matching score, external evidence credibility score and spatial consistency score, realizing the integrated utilization of name information, spatial constraints and external historical materials, overcoming the shortcomings of existing technologies that rely solely on rule matching or local evidence judgment, and making the continuous identification of historical city entities more systematic and comprehensive.

[0047] (4) Improve the expression of subtle displacements in urban centers under small scale conditions: In view of the problem that existing trajectory overlay methods are difficult to clearly present subtle displacement directions and spatial differences on small scale maps, this invention adopts a regional visualization method based on Voronoi units and HSV color coding, which maps displacement direction to hue and displacement intensity to saturation, thereby enabling the expression of regional migration direction structure and displacement intensity differences in a single layer, significantly enhancing the readability and comparability of displacement patterns.

[0048] (5) Capable of supporting the construction and analysis of long-term historical evolution chains: This invention can not only achieve place name matching between adjacent periods, but also further form multi-period connected components and historical urban entity evolution chains, providing continuous data support and analytical foundation for long-term tracking of historical urban entities, regional urban system evolution analysis, and administrative division change research. The results show that the method generates a total of 1487 candidate link decisions, identifies 117 four-period connected components, and ultimately links 767 urban nodes, indicating that the method has good practical applicability.

[0049] (6) It has strong application value: This invention can be applied to fields such as historical geography research, analysis of the spatial evolution of historical cities, research on changes in administrative divisions, and spatiotemporal visualization, providing a practical technical support for the continuous identification of historical city entities and the analysis of urban center migration patterns. Attached Figure Description

[0050] Figure 1 This is an overall flowchart of the method for enhancing the retrieval of cross-period historical place names and visualizing the displacement of urban centers according to the present invention;

[0051] Figure 2 This is a diagram of the city node data structure of the present invention;

[0052] Figure 3 This is a schematic diagram of the network retrieval and LLM scoring of the present invention;

[0053] Figure 4 This is a schematic diagram of the regional displacement visualization of the present invention. Detailed Implementation

[0054] The present invention will be further illustrated below with reference to specific embodiments. These embodiments are implemented based on the technical solutions of the present invention, and it should be understood that these embodiments are only used to illustrate the present invention and are not intended to limit the scope of the present invention.

[0055] To address the issues of place name variants, obsolete names, transliteration differences, spatial ambiguity, and difficulty in clearly representing subtle displacements at small scales in cross-period historical city point data, this embodiment provides a retrieval-enhanced method for linking cross-period historical place names and visualizing urban center displacement. This method is geared towards identifying cross-period historical city entities through place name linking and spatial displacement visualization. It integrates standardized preprocessing, candidate generation, online retrieval, structured evidence extraction based on a large language model, unified scoring, and regionalized displacement representation into a single technical framework, thereby achieving continuous identification of historical city entities and intuitive presentation of urban center migration patterns.

[0056] Combination Figure 1 As shown, the present invention mainly includes the following steps:

[0057] Step 1: Acquisition and standardization preprocessing of historical city center point data;

[0058] Step 101: Collect city center point data from multiple historical periods, and map fields such as place name, coordinates, administrative type and administrative level from the original data of each period to a unified field structure to form a cross-period city node dataset.

[0059] like Figure 1As shown, in this embodiment, urban center point data from four periods in Northwest China are selected as input, specifically including center point data for Chang'an County, Shaanxi Province in 1820 and 1911, and center point data for Xi'an City, Shaanxi Province in 1970 and 2020. The data for 1820 and 1911 are from CHGIS (China Historical Geographic Information System), the data for 1970 is from CORONA (Keyhole Satellite) image interpretation results, and the data for 2020 is from the Tianditu platform and related administrative location data. The field mapping process is represented as follows:

[0060]

[0061] in, Indicates the period to which the data belongs. express The first period One original record, Indicates the applicable period Field mapping functions, This represents a standardized structure record after field mapping is complete.

[0062] In one implementation, the unified fields include at least: period, original data source, original row number, original place name, latitude and longitude, administrative type, provincial-level administrative region, prefecture-level administrative region, and county-level administrative region. For data with inconsistent field names, candidate fields are retrieved sequentially according to a preset priority, and coordinate fields that cannot be converted into numerical values ​​are set to null values.

[0063] Step 102: Preprocess the collected city center point data and output standardized city node data;

[0064] Subsequently, the raw point data undergoes standardization to improve the comparability of data from different periods. Standardization includes at least: name format standardization, administrative level standardization, time label unification, and spatial representation unification. Name standardization includes unifying simplified and traditional Chinese characters, standardizing variant characters, separating administrative suffixes, and generating pinyin fields. Administrative semantic standardization includes unified coding of administrative type, superior affiliation, and central location attributes. This process is represented as:

[0065]

[0066] in, This represents the original text to be processed. This indicates character conversion operations, including conversion from traditional Chinese characters to simplified Chinese characters. This indicates a whitespace cleanup operation, including deleting regular spaces, full-width spaces, and tabs. This indicates a unified operation for punctuation marks. This indicates the deletion of any remaining punctuation marks at the beginning and end of the line. This represents the standardized text.

[0067] Furthermore, the core part of the place name and the administrative division suffix are separated using the longest suffix matching method, specifically:

[0068]

[0069]

[0070] in, Indicates the first Standardized place names in the records This indicates a predefined set of administrative division suffixes, including "county," "city," "district," "prefecture," "department," "banner," "township," "town," "village," and "autonomous county," etc. This indicates the longest administrative division suffix identified from standardized place names. Indicates the length of the suffix string. This refers to the core part of a place name after removing the administrative division suffix.

[0071] Based on the identified administrative division suffixes, city nodes are normalized into administrative categories such as county-level, prefecture-level, township-level, or village-level. Simultaneously, a quality score can be performed on the integrity of the node data, specifically:

[0072]

[0073] in, Indicates the first Data quality score for each node Indicates whether the coordinates are invalid. This indicates whether standardized place names are missing. Indicate whether the administrative type is missing. This indicates whether the provincial-level administrative region field is missing. The indicator function will execute when the corresponding condition is met. The value is 1 otherwise. 0.5, 0.2, 0.1 and 0.05 are the preset deduction weights for each type of missing information.

[0074] Step 103: Spatial attribute association;

[0075] First, the latitude and longitude coordinates of city nodes are validated. In one implementation, for data within the Chinese region, the coordinate validity criteria are as follows:

[0076]

[0077] in, Indicates the first The longitude of each node Indicates the first The latitude of each node Indicates whether the node coordinates are valid. When the latitude and longitude are within the preset range, =1, otherwise .

[0078] The preset range can be adjusted according to the target research area.

[0079] For nodes with valid coordinates, they are converted into spatial point objects in a geographic coordinate system and spatially associated with modern administrative boundaries, specifically as follows:

[0080]

[0081] in, Indicates the first Spatial points corresponding to each city node Indicates the first A polygonal boundary of a modern administrative division. Represents the boundary polygon The internal area, This indicates the name and administrative code of the modern administrative division corresponding to the polygon. Represents a node The attributes of modern administrative divisions obtained through association.

[0082] Furthermore, the spherical distances between historical nodes and each modern reference center point are calculated, that is, the spherical distances between the source node and the target node are calculated based on the node's latitude and longitude. The distance calculation formula is:

[0083]

[0084]

[0085]

[0086] in, and Representing historical nodes Longitude and latitude and They represent the modern reference center points respectively. The longitude and latitude are converted to radians before being used in the calculation. This represents the average radius of the Earth, which is taken as 6371.008 km in one implementation. This represents an intermediate variable in the process of calculating spherical distance. Representing historical nodes With the modern reference center point The spherical distance between them Represents the set of modern reference center points. Indicates historical nodes The nearest modern reference center.

[0087] In one implementation, the modern reference center point is derived from the most recent urban location data. The spatial attribute association result includes at least the modern administrative division name, the modern administrative division code, the name of the nearest modern reference center, and the corresponding distance.

[0088] Step 104, Node Output;

[0089] Records that have been standardized and associated with spatial attributes will be output as unified city nodes. The unique identifier for each node is represented as follows:

[0090]

[0091] in, Indicates the first A unique identifier for each city node Indicates the period to which the node belongs. This indicates the row number of the node in the original data. This means padding the line number to six digits. This indicates a string concatenation operation.

[0092] For example, the node identifier "1820_000001" indicates that the node originates from the first original record in the data from 1820.

[0093] The merged set of nodes is deduplicated, specifically as follows:

[0094] ;

[0095] in, Indicating period The corresponding standardized set of nodes, This represents the result of merging multiple sets of nodes from different periods. Indicates the period to which the node belongs. Indicates standardized place names, and These represent the longitude and latitude of the node, respectively. This indicates a deduplication operation. This represents the standardized set of city nodes in the final output.

[0096] Through the above processing, standardized urban node data with unified field structure, standardized place name expression, clear administrative attributes and spatial correlation information are formed, providing a data foundation for subsequent cross-period candidate node generation, similarity calculation and identification of urban evolution relationships.

[0097] Step 2: Generate candidate place name pairs;

[0098] based on Figure 2 The standardized city node data shown generates candidate place name pairs between adjacent periods or preset period combinations. For example... Figure 1 As shown, the generation of candidate place name pairs includes four steps: spatial attribute and place name attribute input, text similarity scoring and administrative level constraints, spatial constraints and candidate selection, and candidate pair sorting and output. Among them, the multi-condition selection method is mainly applied to the second and third steps, including name similarity selection, administrative compatibility selection, and spatial constraint selection.

[0099] Step 201, Spatial Attribute and Place Name Attribute Input: Obtain standardized city node data for the source period and the target period, and generate an initial set of node pairs to be compared based on the Cartesian product operation of the city node set for the source period and the city node set for the target period;

[0100] Obtain standardized city nodes for the period to be compared, and generate an initial set of node pairs to be compared:

[0101]

[0102] in, Indicates the period of origin The set of city nodes, Indicates the target period The set of city nodes, This represents the Cartesian product operation. This represents the initial set of node pairs to be compared between two periods.

[0103] Input attributes include standardized place name, place name core, place name pinyin, administrative type, provincial-level administrative region, prefecture-level administrative region, county-level administrative region, latitude and longitude, modern administrative division name, nearest modern reference center point, and remarks text.

[0104] In one implementation, the preset period combination includes 1820 to 1911, 1911 to 1970, and 1970 to 2020.

[0105] Step 202, Text Similarity Scoring and Administrative Level Constraints: Compare the place name attributes and administrative attributes of the initial pairs of nodes to be compared; calculate the comprehensive name score based on the similarity of standardized place names, place name core, place name pinyin and character sequence, and calculate the administrative compatibility score by combining administrative type compatibility and administrative context similarity;

[0106] First, the similarity of the place name attributes of the two nodes is compared. Character sequence similarity is represented as:

[0107]

[0108] in, and This represents two text sequences to be compared. This indicates the number of matching characters identified by the sequence matching algorithm. and These represent the lengths of the two text sequences, Represents the normalized character sequence similarity, with values ​​ranging from 1 to 2. .

[0109] The similarity is 0 when either text is empty; the similarity is 1 when the two texts are completely identical.

[0110] Overall Name Score Represented as:

[0111]

[0112] in, This indicates whether the standardized place names are completely identical; a value of 1 indicates identical names, and a value of 0 indicates otherwise. This indicates whether the core meaning of place names remains the same after removing the administrative division suffix. Indicates the similarity of place name pinyin sequences. The character sequence similarity of standardized place names is represented by 0.35, 0.30, 0.15 and 0.20, which are preset weights.

[0113] For example, "Zhuanglang Hall" and "Zhuanglang County" have different administrative suffixes, but the core of the place name is "Zhuanglang", so their potential connection can be preserved.

[0114] Subsequently, based on the administrative type and the superior administrative division, it is determined whether the node pair conforms to the logic of administrative succession:

[0115]

[0116] in, This indicates the overall score for administrative compatibility. This indicates the administrative type compatibility score (determined based on a preset rule table). This indicates the similarity of administrative contexts, with 0.55 and 0.45 being preset weights.

[0117] For periods with similar systems, the administrative context similarity is represented as:

[0118]

[0119] For periods with significant differences in administrative systems, modern spatial anchoring is used to assist in judgment:

[0120]

[0121] in, Indicates the similarity of provincial-level administrative regions. Indicates the similarity of prefecture-level administrative regions. Indicates the similarity of county-level administrative regions. Indicating the similarity of names of modern administrative division boundaries, This indicates the similarity of the names of the most recent modern reference center points.

[0122] Step 203, Spatial Constraints and Candidate Selection: Calculate the spherical distance between the source node and the target node based on the latitude and longitude of the node, and set an adaptive distance threshold based on the combination of node administrative category and period; combine name similarity, administrative compatibility and spatial constraints to jointly select the initial pairs of nodes to be compared to form a set of candidate place name pairs;

[0123] Calculate the spherical distance between the source node and the target node:

[0124]

[0125]

[0126] in, and These represent the longitude and latitude of the source node, respectively. and These represent the longitude and latitude of the target node, respectively. The longitude and latitude are converted to radians before calculation. This represents the average radius of the Earth, which is taken as 6371.0088 km in one implementation. This represents the spherical distance between two nodes. This represents the dimensionless intermediate variable in the Haversine distance calculation.

[0127] Set adaptive distance thresholds for different node types and time periods:

[0128]

[0129] in, This indicates the upper limit of the allowed reference distance. This represents the base distance threshold corresponding to the administrative category. This indicates the adjustment factor corresponding to the period combination.

[0130] Examples of basic distance thresholds are shown in Table 1 below:

[0131] Table 1 Basic Distance Thresholds

[0132]

[0133] Overall Spatial Score Represented as:

[0134]

[0135]

[0136] in, This represents the distance decay score. This indicates whether two nodes are associated with the same modern administrative division boundary; a value of 1 indicates yes, and a value of 0 indicates no. This indicates whether two nodes have the same nearest modern reference center point; if yes, it is set to 1, otherwise it is set to 0. Indicates the upper limit of the allowed reference distance; This represents the spherical distance between two nodes.

[0137] Candidates were selected through a combination of name, administrative, and spatial criteria:

[0138]

[0139] in, This represents the set of candidate place name pairs after filtering. This represents the name similarity threshold. This indicates the threshold for administrative context similarity. This represents the spatial score threshold. Indicates the administrative type compatibility score; Indicates the similarity of administrative contexts; This represents the overall score for name similarity. This represents the overall score of spatial attributes. Indicates the period of origin and target period The standardized city nodes are generated by the Cartesian product to form the initial set of node pairs to be compared.

[0140] Step 204, Candidate Pair Sorting and Output: Calculate the first-stage comprehensive score based on the comprehensive name score, administrative compatibility score, spatial score, and on-site description text hit rate; for the same source node, sort the candidate place name pairs in descending order according to the first-stage comprehensive score, and output the top K candidate results to provide candidate data basis for subsequent network retrieval, evidence extraction, and cross-period place name link determination.

[0141] Calculate the first-stage comprehensive score for the candidate place names that have passed the screening:

[0142]

[0143] in, This represents the first-stage overall score for the candidate place name pair. This represents the overall score for name similarity. This indicates the overall score for administrative compatibility. This represents the overall score of spatial attributes. This indicates the hit rate of the remarks text or on-site description text to the target place name and administrative attributes. 0.42, 0.25, 0.23 and 0.10 are preset weights.

[0144] For nodes from the same source, candidates are ranked from highest to lowest based on their overall score, and the top candidates are retained. One result:

[0145]

[0146] in, Indicates the source node The corresponding final set of candidate place name pairs, This indicates that the scores are retained in descending order based on the overall score. result, This represents the maximum number of candidates that each source node can retain, and in one implementation, it is set to 8. This represents the first-stage overall score for the candidate place name pair. This represents the set of candidate place name pairs after filtering.

[0147] The unique identifier of a candidate pair is represented as:

[0148]

[0149] in, A unique identifier representing a pair of candidate place names. and These represent the identifiers of the source node and the target node, respectively. This indicates a string concatenation operation.

[0150] Through the above processing, a set of candidate place name pairs is obtained that takes into account textual similarity, administrative inheritance rationality, and spatial location rationality. This provides a candidate data foundation for subsequent network retrieval, evidence extraction, credibility evaluation, and place name association determination. The city node data structure is as follows: Figure 2 As shown.

[0151] Step 3: Enhanced evidence acquisition and structured analysis of candidate place name pairs;

[0152] For candidate place name pairs that have undergone significant name changes, administrative type changes, or cannot be reliably determined solely by rule matching, evidence enhancement processing should be performed. For example... Figure 1 and Figure 3 As shown, the process includes web retrieval enhancement, structured extraction from large language models, and evidence credibility scoring.

[0153] Step 301, Enhanced Web Search:

[0154] First, candidates for web search are selected from the set of candidate place name pairs. The search trigger condition is expressed as follows:

[0155]

[0156] in, Indicates the candidate place name pair Whether to trigger network search This indicates the overall score for the first stage. This indicates the name similarity score. This represents the overall score of spatial attributes. Represents the semantic similarity of optional word vectors. This indicates whether there are complex changes in administrative type, such as "a department changing to a county" or "a county changing to a district". This indicates whether there are multiple high-scoring candidates for the same source node. and These represent the lower and upper limits of the retrieval trigger interval, respectively. This indicates an indicator function that takes the value 1 if the condition within the parentheses is true, and 0 otherwise.

[0157] In one implementation, 0.50, 0.85. When there are at least two candidates with a composite score of not less than 0.72 from the same source node, let .

[0158] For candidate place name pairs that trigger a search, extract the source place name, target place name, administrative type, provincial-level administrative region, prefecture-level administrative region, on-site description, and remarks text, and automatically construct a set of query statements:

[0159]

[0160] in, This represents the set of query statements corresponding to each candidate location name pair. and These represent the source and destination place names, respectively. Indicates the prefecture-level administrative region to which the source node belongs. Indicates the provincial-level administrative region to which the source node belongs. and These represent the administrative types of the source node and the target node, respectively. This represents the local description text of the source node. The comment text representing the source node. This indicates the deduplication operation for the query statement.

[0161] When the source place name is the same as the target place name, you can construct query statements such as "source place name + history", "source place name + present location" and "source place name + administrative division".

[0162] Based on the triggering reasons of the candidate options, the search source types are prioritized. See Table 2 below for details.

[0163] Table 2 Candidate Triggering Reasons Table

[0164]

[0165] It should be noted that source type priority is used to organize retrieval tasks and label evidence sources. In one implementation, search results are obtained through a general online retrieval interface, and each result includes at least a title, URL, and abstract text.

[0166] Step 302: Structured evidence extraction based on a large language model:

[0167] The query statement and its corresponding prefix Inputting the network search results into the large language model:

[0168]

[0169] in, This represents a query statement. Indicates the first The title of each search result. Indicates the first The URLs of the search results Indicates the first A summary of the search results. This indicates the number of search results retained for each query statement; in one implementation, it is set to 5. This represents the structured analysis process of a large language model. This represents the structured text output by the model.

[0170] The structured text should include at least: place name identification results; the historical period to which the place name belongs; the administrative level and region to which it belongs; alternative names, name changes and changes in administrative divisions; the relocation of the administrative center and clues to historical inheritance; similarity measurement results; and a summary of key evidence.

[0171] The semantic similarity output by the large language model is denoted as:

[0172]

[0173] in, This represents the similarity score output by the large language model, with a value ranging from 0 to 100. Represents the normalized semantic similarity, with values ​​ranging from 1 to 10. .

[0174] When the model output cannot be parsed, you can Set to the default value of 0.5.

[0175] Furthermore, weights can be analyzed based on semantic similarity calculation models:

[0176]

[0177] in, This represents the weights corresponding to the analysis results of the large language model. 0.3 indicates the minimum retained weight, and 0.6 indicates the weight range that increases with semantic similarity. The range of values ​​is .

[0178] In one alternative implementation, the large language model may also be required to output candidate pair scoring results in JSON format, including candidate pair identifier, source location name, target location name, semantic score, decision type, decision reason, key evidence, and remarks. Decision types include match, likely match, uncertain, likely mismatch, and mismatch.

[0179] Step 303: Evidence credibility scoring;

[0180] For the same candidate place name, multiple pieces of evidence are aggregated using a weighted method. First, source quality weights are set based on the URLs of the evidence sources, as shown in Table 3 below.

[0181] Table 3 Sources of Evidence

[0182]

[0183] For the The evidence is analyzed using keywords related to support, conflict, and uncertainty in the structured text to calculate an evidence propensity score, specifically:

[0184]

[0185] in, Indicates the first The tendency score of each piece of evidence, This indicates whether the evidence contains supporting keywords such as "support," "prove," or "consistent." Indicate whether the evidence contains conflicting keywords such as "contradictory," "different," or "inconsistent." This indicates that the results will be restricted to... Within the range.

[0186] When evidence contains vague keywords such as "possible," "suspected," or "uncertain," Set it to 0.5.

[0187] The overall score for a single piece of evidence is expressed as follows:

[0188]

[0189] in, Indicates the first The overall score of each piece of evidence, Indicates the first The normalized semantic similarity corresponding to each piece of evidence. Indicates the first The tendency score for each piece of evidence is set with preset weights of 0.7 and 0.3.

[0190] The credibility score of the evidence for candidate place names is expressed as follows:

[0191]

[0192] in, The score indicates the credibility of the evidence for each candidate place name pair. This indicates the amount of evidence corresponding to the candidate place name. Indicates the first The source quality weight of each piece of evidence. Indicates the first The overall score of each piece of evidence.

[0193] Confidence levels are determined based on the credibility score of the evidence:

[0194]

[0195] Through the above processing, unstructured historical texts are transformed into quantifiable matching evidence, providing a basis for subsequent place name association determination. Indicates the confidence level of the evidence for the candidate place name pair; , and These represent high confidence, medium confidence, and low confidence levels, respectively. This indicates the credibility score of the evidence for each candidate place name pair.

[0196] In this embodiment, the online retrieval tool uses the Serper interface, but can be replaced with the SerpAPI or Bing interface. The large language model uses the Claude series model, and the model identifier in the configuration example is claude-opus-4-6. The structured analysis includes at least the alias relationships, renaming relationships, administrative center migration information, administrative characteristics, and historical inheritance clues between candidate place name pairs, and outputs semantic matching scores and a summary of key evidence.

[0197] Step 4: Standardized scoring and determination of cross-period place name links;

[0198] The first-stage comprehensive score, external evidence credibility score, and spatial consistency score of candidate place name pairs are incorporated into a unified scoring framework to determine the link between candidate place name pairs.

[0199] First, limit the scores of each input item to a range. Inside:

[0200]

[0201] in, This represents the raw score to be processed. This represents the score limiting function.

[0202] The final overall score is expressed as follows:

[0203]

[0204] in, This represents the final overall score for the candidate place name pairs. This represents the first-stage comprehensive score obtained during the candidate pair generation phase. This represents the credibility score of the evidence obtained based on external search results and large language model analysis. The score represents the spatial consistency score. 0.45, 0.35 and 0.20 represent the preset weights corresponding to the three scores, and the sum of the weights is 1.

[0205] Step 401, First Stage Overall Score calculate;

[0206] The comprehensive evaluation result at the rule matching level is calculated during the candidate pair generation process in step 2.

[0207]

[0208]

[0209]

[0210]

[0211] in, This represents the overall score for name similarity. This indicates the overall score for administrative compatibility; This represents the spatial constraint score of the candidate pair during the generation phase. This indicates whether the on-site description or remarks text matches the target place name and related administrative attributes. 0.42, 0.25, 0.23 and 0.10 are the preset weights.

[0212] Step 402, External Evidence Credibility Score calculate;

[0213] For candidate place name pairs that perform retrieval enhancement, multiple external evidences corresponding to the same candidate pair are weighted and aggregated:

[0214]

[0215]

[0216] in, Indicates the evidence number, This indicates the number of pieces of evidence corresponding to each candidate place name. Indicates the first The normalized semantic similarity corresponding to each piece of evidence. Indicates the first The tendency score of each piece of evidence, Indicates the first The overall score of each piece of evidence, Indicates the first The source quality weight of each piece of evidence. This represents the credibility score of external evidence for candidate place name pairs.

[0217] When no valid evidence is available, the scoring module will... Set to 0. The main process can generate a neutral placeholder score even when no retrieval is performed and no available evidence scoring file exists. This is to support the operation of subsequent processes.

[0218] Step 403, Spatial Consistency Score calculate;

[0219] For candidate place name pairs Taking into account factors such as distance between nodes, administrative type, overlapping relationships of modern administrative divisions, and similarity of place name characters, the spatial consistency score is calculated:

[0220]

[0221] in, Indicates candidate nodes and Spatial consistency score, This represents the score based on the distance between nodes. Indicates the similarity of administrative types. This indicates the score for overlapping modern administrative divisions. Indicates the similarity of place name characters or lexical units. This represents a score limiting function that restricts the score to a certain value. Within the interval, 0.50, 0.20, 0.20, and 0.10 are preset weights.

[0222] First, the Haversine formula is used to calculate the spherical distance \(d\) between two nodes, and a distance score is generated based on the distance:

[0223]

[0224] Wherein, represents the spherical distance between two nodes, with the unit being km.

[0225] The administrative type similarity is determined according to preset rules:

[0226]

[0227] Wherein, when , it indicates that the administrative type is missing; when , it indicates that the administrative types are the same; when , it indicates that the administrative macro categories are the same; when , it indicates that the prefixes of the administrative types are the same; when , it represents other situations.

[0228] The modern administrative division overlap score is determined according to the following rules:

[0229]

[0230] Wherein, when , it indicates that the modern administrative division codes are the same; when , it indicates that the modern administrative division names are the same; when , it indicates that the administrative divisions are different but the distance between nodes does not exceed 5 km; when , it represents other situations.

[0231] The place name character or token similarity is calculated using the Jaccard coefficient:

[0232]

[0233] Wherein, and respectively represent the character or token sets corresponding to the place names of two nodes. For Chinese place names, segmentation is performed by character; for text containing spaces, segmentation is performed by token.

[0234] When any node is missing, the spatial consistency score is set to a neutral default value of 0.5. Through the above processing, the resulting value range is... Spatial consistency score .

[0235] Step 404: Determine the link result;

[0236] Based on the final comprehensive score, the candidate place name pairs are divided into three categories: automatically linked, manually reviewed, and automatically rejected.

[0237]

[0238] in, This indicates that cross-period place name links will be automatically created. This indicates that the candidate place names will be added to the list for manual review. This indicates that the connection will be automatically rejected. Indicates the candidate place name pair The link determination result; This indicates the final overall score of the candidate place name pair.

[0239] The list of manually reviewed documents is arranged in descending order of the final overall score:

[0240]

[0241] in, This indicates the list of items to be manually reviewed. This indicates that the scores are sorted from highest to lowest according to the specified scores. Indicates the candidate place name pair The link determination result; This indicates the final overall score for the candidate place name pair; This indicates that the candidate place name pair has entered the manual review stage.

[0242] Through the above processing, candidate place name pairs are automatically linked, manually reviewed, or automatically rejected, and the results of cross-period place name linking are output, along with a list of manually reviewed names sorted in descending order of the final comprehensive score. For candidates undergoing enhanced retrieval, the external evidence credibility score can reflect clues such as aliases, name changes, relocation of administrative centers, and changes in administrative divisions contained in web evidence and large language model analysis, thereby reducing the limitations of relying solely on name similarity and spatial distance for judgment. The confirmed cross-period linking results can serve as the data foundation for subsequently constructing long-term evolutionary chains.

[0243] Step 5: Construction of urban center displacement samples;

[0244] Based on the cross-period place name linking results obtained in step 4, adjacent period node pairs belonging to the same historical city entity are extracted. Adjacent periods include 1820-1911, 1911-1970, and 1970-2020. For the... There are 10 valid link records, with the starting and ending coordinates as follows:

[0245]

[0246] in, Indicates the city center point at the time of origin. Indicates the city center point during the target period. and These represent the longitude and latitude of the node, respectively.

[0247] Calculate the displacement vector of the city center:

[0248]

[0249]

[0250] in, Indicates the first Each link records the corresponding city center displacement vector; and These represent the coordinate differences between the endpoint and the starting point in the X and Y directions, respectively. Indicates the longitude for the target period; Indicates the longitude of the period of origin; Indicates the latitude of the target period; Indicates the latitude of the period of origin.

[0251] The Euclidean length of the displacement vector is used as the weight of the displacement sample:

[0252]

[0253] in, Indicates the first Each link records the corresponding displacement length. This represents the weight of the displacement sample. This weight is calculated from the city center displacement and does not use the comprehensive score in the cross-period linkage determination. .

[0254] Further calculate the direction angle from the starting point to the ending point:

[0255]

[0256] Normalize the direction angle to the interval :

[0257]

[0258] in, This represents the normalized displacement direction angle. Direction angles of 0°, 90°, 180°, and 270° represent displacements to the east, north, west, and south, respectively. This represents the original square angle pointing from the starting point to the ending point.

[0259] For valid displacement records, the midpoint between the start and end points is taken as the spatial sample location for subsequent regional visualization:

[0260]

[0261] in, Indicates the first The spatial representative point of each displacement sample is the midpoint of the line connecting the starting point and the ending point.

[0262] According to the preset tolerance The linked records are divided into fixed points and effective displacement samples:

[0263]

[0264] in, Fixed points represent city nodes where the starting and ending points approximately coincide, and serve as fixed generation points for subsequent spatial partitioning; effective displacement samples include sample number, source link number, midpoint coordinates, orientation angle, and displacement weight. Indicates the first The link records the sample type obtained after conversion; This represents a fixed-point sample where the starting and ending points approximately coincide. This represents a displacement sample with an effective displacement direction and displacement weight.

[0265] Through the above processing, the cross-period urban link results are transformed into a set of urban center displacement samples that can be used for subsequent spatial partitioning and directional visualization.

[0266] Step 6: Visualize the regional displacement based on the Voronoi element;

[0267] To address the difficulty in directly identifying subtle displacements on small-scale maps, a regionalized displacement representation method based on Voronoi cells and HSV color coding is adopted. This includes the following steps.

[0268] Step 601: Construct Voronoi elements and aggregate displacement samples;

[0269] The effective displacement samples obtained in step 5 are used as discrete space samples. Let the study area be... , No. The number of generated points is The corresponding Voronoi element is:

[0270]

[0271] in, Indicates the first One Voronoi unit, This represents any location within the study area. and This represents the Voronoi generation point. All generated cells are clipped to the study region. Inside.

[0272] This implementation uses the Lloyd iterative method with fixed generation point constraints to optimize Voronoi elements. For elements containing valid displacement samples, the generation point positions are updated according to the sample weights:

[0273]

[0274] in, Indicates the first The generated point after the next iteration Indicates allocation to unit The sample set, Indicates the midpoint of the sample coordinates This represents the displacement weights calculated in step 5. For the generated points corresponding to fixed points, no position update is performed; for cells not assigned to valid samples, the original generated point positions are retained.

[0275] Iteration stops when the maximum distance the generated point moves is less than a preset threshold or the maximum number of iterations is reached.

[0276]

[0277] In one implementation, the movement distance threshold The maximum number of iterations is 25, and the number of normally generated points is 80. The fixed points obtained in step 5 are added in addition to these. Indicates the first The Voronoi generation point is at the _ ... Spatial position after Lloyd iteration.

[0278] After partitioning, each displacement sample is assigned to its corresponding Voronoi element. In one implementation, assignment is prioritized based on spatial inclusion; for samples at boundary locations, assignment is based on spatial intersection; if assignment still fails, the sample is assigned to the nearest element. For each element, the number of samples, the sum of sample weights, and the sample weight density are recorded.

[0279]

[0280] in, Indicates the number of samples within a cell. This represents the sum of sample weights within a unit. Represents the area of ​​a unit. This represents the sample weight density. The above statistical values ​​are used to record unit features; the subsequent HSV saturation uses the sample weight mean. This represents the displacement weight calculated in step 5. Indicates allocation to unit The sample set.

[0281] Step 602: Calculate the dominant direction of each Voronoi element using a weighted circumferential average value;

[0282] Because orientation angles are periodic—for example, the average orientation of 359° and 1° should be close to 0°—the ordinary arithmetic mean cannot be directly used. For a given unit... For each valid displacement sample within the range, calculate the weighted cosine sum, weighted sine sum, and weighted sum:

[0283]

[0284] in, Consistent with the definition in step 5, it indicates that the first The normalized orientation angle of each displacement sample; Indicates displacement weight; This represents the weighted cosine sum; This represents the weighted sine sum. Superscript is used. This is to distinguish it from the saturation component in the HSV model.

[0285] Calculate the dominant orientation angle of the unit:

[0286]

[0287] Normalize it to an interval :

[0288]

[0289] Further conversion to angle system:

[0290]

[0291] in, This represents the original dominant orientation angle of the k-th Voronoi element, calculated based on the weighted sum of sine and weighted sum of cosine. This represents the normalized dominant orientation angle of the unit; This indicates the dominant direction angle after conversion to a degree system.

[0292] Step 603: Map the dominant direction to the hue components in the HSV color model;

[0293] Mapping the dominant orientation angle of the unit to the hue component in the HSV color model:

[0294]

[0295] in, Indicates the first The hue component of each Voronoi unit has a value range of 1. Since the directional angle maps back to the same hue after 360°, it is possible to maintain the continuity and periodicity of the directional variable.

[0296] Step 604: Map the average weight of samples within the cell to the saturation component in the HSV color model;

[0297] First, calculate the average weight of the effective displacement samples within each Voronoi element:

[0298]

[0299] in, Indicates the first The original displacement strength of each element. Indicates the first The number of valid displacement samples within a Voronoi element; Indicates allocation to unit The sample set, This represents the displacement weight.

[0300] Subsequently, the minimum value was calculated among all effective elements with original displacement strength greater than 0. and maximum value :

[0301]

[0302]

[0303] when When using minimum-maximum value normalization:

[0304]

[0305] in, Represents normalized displacement intensity. This indicates that the calculation result is restricted to an interval. Internally. When the original displacement strength of all effective elements is the same, the effective element's... Set it to 0.5; for elements with no effective displacement samples, set it to 0.

[0306] Mapping normalized displacement intensity to HSV saturation:

[0307]

[0308] in, Indicates the first The HSV saturation of each element is determined by the normalized displacement intensity. and These represent the lower and upper limits of the HSV saturation mapping, respectively, used to control the minimum and maximum values ​​of color saturation.

[0309] Step 605: Generate the orientation-intensity joint color for each Voronoi unit;

[0310] Keep the luminance component in the HSV color model constant:

[0311]

[0312] Finally, the color of each Voronoi unit is represented as:

[0313]

[0314] The generated HSV colors are converted to hexadecimal color values ​​and written to the dir_hex field for Voronoi unit shading. Hue reflects the dominant displacement direction of the region, saturation reflects the intensity of the displacement, and brightness remains unchanged, thus expressing both the directional structure and intensity differences of region migration in a single layer. This represents the luminance component in the color model and is set as a constant.

[0315] Step 7: Output and Application of Results

[0316] It outputs cross-period historical place name links, historical city entity evolution chains, and visualizations of city center displacement. The results can be used for applications such as continuous identification of historical city entities, regional urban system evolution analysis, administrative division evolution, and long-term city center migration pattern recognition.

[0317] This embodiment also provides a preferred implementation. Taking five provincial-level regions in Northwest China—Shaanxi, Ningxia Hui Autonomous Region, Qinghai, Gansu, and Xinjiang Uygur Autonomous Region—as the research object, urban center point data from four periods—1820, 1911, 1970, and 2020—are input. First, name and administrative semantic standardization processing is performed. Then, candidate place name pairs are generated through name similarity, spatial distance threshold, and administrative compatibility. For candidates that are difficult to determine directly, historical evidence is obtained through online retrieval, and information on aliases, administrative center migration, and administrative inheritance is extracted using Claude Opus 4.6. After calculating a comprehensive score, the cross-period place name link results are determined. Furthermore, displacement trajectories are constructed based on successfully matched urban entities, and Voronoi units and HSV color coding are used to express their dominant migration direction and displacement intensity, thereby achieving regionalized visual analysis of historical urban center migration patterns.

[0318] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for enhancing the retrieval of cross-period historical place name links and visualizing urban center displacement, characterized in that: Includes the following steps: Step 1: Acquisition and standardization preprocessing of historical city center point data; Step 2: Based on the standardized city node data, generate candidate place name pairs between adjacent periods or preset period combinations; Step 3: Enhanced evidence acquisition and structured analysis of candidate place name pairs; Step 4: Standardized scoring and determination of cross-period place name links; Step 5: Construction of urban center displacement samples; Step 6: Visualize the regional displacement based on the Voronoi element; Step 7: Output and Application of Results 2. The method for enhanced retrieval of cross-period historical place name links and visualization of urban center displacement according to claim 1, characterized in that: In step 1, the specific implementation process is as follows: Step 101: Collect city center point data from multiple historical periods to form a cross-period city node dataset; Step 102: Preprocess the collected city center point data and output standardized city node data; Step 103, Spatial Attribute Association: Verify the validity of the latitude and longitude of the city nodes. For nodes with valid coordinates, convert them into spatial point objects in the geographic coordinate system and associate them with the boundaries of modern administrative divisions. Step 104, Node Output: Output the records that have been standardized and associated with spatial attributes as unified city nodes.

3. The method for enhancing the retrieval of cross-period historical place names and visualizing urban center displacement according to claim 1, characterized in that: In step 2, the specific implementation process is as follows: Step 201: Obtain standardized city node data for the source and target periods, and generate an initial set of node pairs to be compared using Cartesian product operation; Step 202: Compare the place name attributes and administrative attributes of the initial node pairs to be compared. Calculate the comprehensive name score based on the similarity of standardized place names, place name core, place name pinyin, and character sequence. Calculate the administrative compatibility score by combining administrative type compatibility and administrative context similarity. Step 203: Calculate the spherical distance between the source node and the target node based on the node's latitude and longitude, and set an adaptive distance threshold based on the combination of node administrative category and period; By combining name similarity, administrative compatibility and spatial constraints, the initial pairs of nodes to be compared are jointly screened to form a set of candidate place name pairs; Step 204: Calculate the first-stage comprehensive score based on the comprehensive name score, administrative compatibility score, spatial score, and the hit rate of the on-site description text; for the same source node, sort the candidate place name pairs in descending order according to the first-stage comprehensive score, and output the top K candidate results to provide candidate data basis for subsequent network retrieval, evidence extraction, and cross-period place name link determination.

4. The method for enhancing the retrieval of cross-period historical place name links and visualizing urban center displacement according to claim 1, characterized in that: In step 3, the specific implementation process is as follows: Step 301, Enhanced Web Search: Conduct web searches from multiple sources and construct search tasks, performing search queries based on source / target location name attributes; Step 302: Extract the structured data from the query statement using the Claude large language model; Step 303: Score the confidence level of evidence based on the consistency of multi-source data, the authority weight of the source, and whether there is a conflict.

5. The method for enhancing the retrieval of cross-period historical place names and visualizing urban center displacement according to claim 1, characterized in that: In step 4, the specific implementation process is as follows: First, the semantic matching results, external evidence credibility, and spatial consistency of candidate place name pairs are incorporated into a unified scoring framework to comprehensively evaluate and score the candidate place name pairs. The formula is: ； in, The semantic matching score is obtained based on the retrieval results and analysis of the large language model. The credibility score for external evidence. Spatial consistency score; Then, the candidate place name pairs are sorted and judged according to the comprehensive score, the cross-period place name link results are determined, and multi-period connected components are further constructed to form a long-term evolution chain of historical city entities.

6. The method for enhancing the retrieval of cross-period historical place name links and visualizing urban center displacement according to claim 1, characterized in that: In step 5, the specific implementation process is as follows: Based on the established cross-period place name linking results, time-series connections are made between different time nodes belonging to the same historical city entity to form city center displacement samples; each displacement sample contains at least the starting point coordinates, ending point coordinates, displacement direction angle and corresponding weight information, thus providing input data for subsequent regional displacement visualization.

7. The method for enhanced retrieval of cross-period historical place name links and visualization of urban center displacement according to claim 1, characterized in that: In step 6, the specific implementation process is as follows: Step 601: Aggregate the original displacement samples into Voronoi elements and record the orientation angle of each sample. and weight ; Step 602: Calculate the dominant direction of each Voronoi element using the weighted circumferential average value; Step 603: Map the average orientation angle to the hue component in the HSV color model to represent the migration direction; Step 604: Normalize the average weight of samples in each unit and map it to the saturation component in the HSV color model to represent the displacement intensity. Step 605: Finally, the color of each Voronoi cell is determined by the hue corresponding to the dominant direction and the saturation corresponding to the displacement intensity; for cells that lack valid samples or cannot obtain valid dominant directions, a no-data color is assigned.

Images