A vector data topological relation automatic checking management system

Abstract

The application discloses a kind of vector data topological relation automatic check management system, belong to GIS data processing technical field, this system includes data access module, topological rule management module, intelligent check engine, automatic repair module, whole-process management module and visual interactive module, by establishing the three-level rule system of " basic rule base+semantic rule base+custom rule", using the dual-engine check architecture of "ontology reasoning+space calculation", the automatic topological check of multi-source vector data is realized, system can automatically detect line segment self-intersection, face element overlap, semantic conflict and other topological errors, and error correction is completed by hierarchical repair mechanism.Simultaneously, the traceability and quality evaluation of whole process are realized by combining blockchain technology, the defects of traditional check in efficiency, accuracy and adaptability are effectively solved, so that the checking time of large-scale vector data is shortened, and the accuracy of error detection is improved.

Description

Technical Field

[0001] This invention belongs to the field of data processing technology in Geographic Information Systems (GIS), specifically relating to an automatic verification and management system for vector data topology relationships. Background Technology

[0002] Vector data is a core data type in GIS systems, using geometric elements such as points, lines, and polygons to accurately represent the location, shape, and attribute information of geographic entities. The accuracy of its topological relationships (such as adjacency, association, and containment) directly impacts the effectiveness of spatial analysis and decision support. In fields such as urban planning, land resource management, and transportation network analysis, inaccurate topological relationships (e.g., polygon overlap, self-intersection of line features, gaps between polygon features) can not only cause data fusion failures but may also lead to biased analysis results and incorrect decisions.

[0003] Current technologies for topological verification of vector data generally suffer from the following problems: First, traditional verification methods rely heavily on manual editing or semi-automated tools, resulting in extremely low processing efficiency. When faced with large-scale, heterogeneous vector data, they struggle to meet timeliness requirements. For example, manual topological checking of large-scale urban land use data can take weeks, with human error often exceeding 15%. Second, the rigidity of verification rules is often based on single geometric features, lacking the ability to adapt to multi-source data fusion scenarios and failing to detect deep logical errors (such as "river segments and lake surfaces should be adjacent rather than intersecting"). Third, the lack of a comprehensive management mechanism leads to a disconnect between verification, repair, and traceability stages, resulting in untraceable modification records of erroneous data, hindering continuous data quality management. Fourth, existing algorithms lack robustness when processing complex data (such as noisy 3D vector data and vector data derived from massive point clouds), with detection accuracy typically below 80%.

[0004] Although GIS platforms such as ESRI and SuperMap have integrated basic topology analysis functions, they have not yet effectively solved the core problems of "low efficiency, rigid rules, scattered management, and poor robustness". Summary of the Invention

[0005] The purpose of this invention is to provide an automatic verification and management system for vector data topology relationships to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: An automatic verification and management system for vector data topology relationships includes a data access module, a topology rule management module, an intelligent verification engine, an automatic repair module, a full-process management module, a visualization interaction module, and a rule optimization module. The modules interact with each other through a data bus. The data access module is used to receive multi-source vector data, complete format parsing, coordinate transformation, spatial index construction and noise filtering preprocessing. During the preprocessing process, the feature extraction unit mines the geometric complexity and attribute correlation of the vector data, and stores the preprocessed data and feature information into a distributed database. The topology rule management module constructs a three-level rule system of "basic rule library + semantic rule library + custom rules". The basic rule library contains 12 types of general topology constraint rules based on the Egenhofer nine intersection model. The semantic rule library combines domain ontology and SWRL rule language to define industry-specific constraints. The custom rule unit supports visual drag-and-drop configuration of rule parameters and logical associations. The intelligent verification engine adopts a three-engine architecture of "ontology reasoning + spatial computing + conflict arbitration". The ontology reasoning engine integrates the Pellet inference engine to realize semantic consistency verification. The spatial computing engine uses an improved edge matching algorithm combined with KD-Tree spatial index to realize geometric topology error detection. The conflict arbitration engine performs probability analysis on the dual-engine verification results and historical data based on the Bayesian network model to solve the contradiction problem of "semantically valid but geometrically conflicting", and finally generates a verification report with confidence label. The automatic repair module performs hierarchical repair based on the error type and confidence level of the verification report. Simple errors are automatically repaired using pruning and buffer fusion methods, while complex errors generate multi-dimensional candidate solutions and quantify the repair impact through a cost-benefit model. After user confirmation, the repair operation is executed, and the repaired data needs to be verified a second time. The full-process management module records traceability information throughout the entire lifecycle using blockchain technology, constructs a four-dimensional quality assessment model of "completeness-accuracy-consistency-timeliness", and supports scheduled and triggered task scheduling and report archiving and export. The visualization and interaction module provides GIS map overlay display, error location, rule configuration panel and quality dashboard, and supports multi-role permission management based on RBAC model; The rule optimization module collects multi-scenario verification data through a federated learning framework and combines reinforcement learning algorithms to dynamically iteratively optimize the parameter thresholds and logical weights of the topology rules. The optimized rules are then manually reviewed and updated to the rule base.

[0007] Preferably, the feature extraction unit of the data access module generates data feature labels by calculating geometric complexity (number of edges of surface features / number of vertices of line features) and analyzing attribute correlation (matching rate of multiple source attributes of the same entity). The coordinate transformation component supports automatic conversion of multiple reference systems such as WGS84 and the 2000 National Geodetic Coordinate System and custom reference system configuration. The spatial index adopts a hybrid index structure of KD-Tree and R-tree to improve the retrieval efficiency of multiple types of features.

[0008] Preferably, the spatial computing engine of the intelligent verification engine adopts the Spark distributed computing framework to achieve sharded parallel processing for tens of millions of vector data. The geometric operation includes six types of error detection: self-intersection, non-closure, overlap, gap, and dangling point. The Bayesian network model of the conflict arbitration engine is trained with historical verification data and has an accuracy rate of no less than 92% in judging contradictory results.

[0009] Preferably, the cost-benefit model of the automatic repair module quantifies the repair solution from three dimensions: data accuracy retention rate, business logic compliance, and repair time. The automatic repair response time for simple errors is ≤500ms, the generation time for candidate solutions for complex errors is ≤3s, and the secondary verification pass rate must reach 98% or higher before the repair process can be completed.

[0010] Preferably, the traceability information of the full-process management module includes data source, verification time, repair operator, rule version, and error confidence. The weights of each dimension in the four-dimensional quality assessment model support industry customization, and the quality score forms a trend comparison curve with historical data.

[0011] Preferably, the federated learning framework of the rule optimization module adopts a horizontal federation mode, which aggregates multi-node verification data under the premise of protecting data privacy. The reinforcement learning algorithm uses "improved verification accuracy + reduced false negative rate" as the reward function. The rule iteration cycle can be configured to 7-30 days. After each iteration, the rule base update log is synchronized to the blockchain.

[0012] Preferably, a vector data topology verification method includes the following steps: S1: The data access module receives multi-source vector data, completes preprocessing and feature extraction, and stores the data and feature labels into a distributed database; S2: The topology rule management module loads the three-level rules, and the rule optimization module calls historical data to pre-optimize the rules; S3: The intelligent verification engine initiates a three-engine collaborative verification process, the conflict arbitration engine handles contradictory results, and generates a verification report with confidence level. S4: The automatic repair module performs hierarchical repair based on the verification report. For complex errors, a cost-benefit model is used to assist in decision-making. After repair, the data is verified a second time. S5: The full-process management module records traceability information and generates a four-dimensional quality score, and archives data and reports; S6: The rule optimization module aggregates the verification data, completes rule iteration through reinforcement learning, and updates the rule base.

[0013] Compared with the prior art, the present invention provides an automatic verification and management system for vector data topology relationships, which has the following beneficial effects: Through automated verification and distributed parallel processing, the verification time for large-scale vector data (such as 1 million line features) has been reduced from the traditional 72 hours to less than 2 hours, improving efficiency by 36 times. The dual-engine architecture of "ontology reasoning + spatial computing" combined with cross-validation mechanism improves the accuracy of error detection to over 95%, significantly reducing the occurrence of false positives and false negatives. The three-tier rule system supports general and industry-specific scenarios, and the custom rule function meets personalized needs, adapting to applications in multiple fields such as land, transportation, and municipal administration. By establishing a full-process traceability and quality assessment mechanism, the problem of "no record of data modification and no quantification of quality" in traditional verification is solved, which helps to continuously optimize data quality. Attached Figure Description

[0014] Figure 1 : Overall system architecture diagram of the present invention. Detailed Implementation

[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] This invention provides, for example Figure 1 shown An automatic verification and management system for vector data topology relationships includes a data access module, a topology rule management module, an intelligent verification engine, an automatic repair module, a full-process management module, a visualization interaction module, and a rule optimization module. The modules interact with each other through a data bus. The data access module is used to receive multi-source vector data, complete format parsing, coordinate transformation, spatial index construction and noise filtering preprocessing. During the preprocessing process, the feature extraction unit mines the geometric complexity and attribute correlation of the vector data, and stores the preprocessed data and feature information into a distributed database. The topology rule management module constructs a three-level rule system of "basic rule library + semantic rule library + custom rules". The basic rule library contains 12 types of general topology constraint rules based on the Egenhofer nine intersection model. The semantic rule library combines domain ontology and SWRL rule language to define industry-specific constraints. Custom rule units support visual drag-and-drop configuration of rule parameters and logical associations. The intelligent verification engine adopts a three-engine architecture of "ontology reasoning + spatial computing + conflict arbitration". The ontology reasoning engine integrates the Pellet inference engine to realize semantic consistency verification. The spatial computing engine uses an improved edge matching algorithm combined with KD-Tree spatial index to realize geometric topology error detection. The conflict arbitration engine performs probabilistic analysis on the verification results of the two engines and historical data based on the Bayesian network model to solve the contradiction problem of "semantically valid but geometrically conflicting", and finally generates a verification report with confidence label. The automatic repair module performs hierarchical repair based on the error type and confidence level in the verification report. Simple errors are automatically repaired using pruning and buffer fusion methods, while complex errors generate multi-dimensional candidate solutions and quantify the repair impact using a cost-benefit model. After user confirmation, the repair operation is executed, and the repaired data needs to be verified a second time. The end-to-end management module uses blockchain technology to record traceability information throughout the entire lifecycle, constructs a four-dimensional quality assessment model of "completeness-accuracy-consistency-timeliness", and supports scheduled and triggered task scheduling and report archiving and export. The visualization and interaction module provides GIS map overlay display, error location, rule configuration panel and quality dashboard, and supports multi-role permission management based on RBAC model; The rule optimization module collects multi-scenario verification data through a federated learning framework and combines reinforcement learning algorithms to dynamically iterate and optimize the parameter thresholds and logical weights of the topology rules. The optimized rules are then manually reviewed and updated to the rule base.

[0017] The feature extraction unit of the data access module generates data feature labels by calculating geometric complexity (number of edges of polygon features / number of vertices of line features) and analyzing attribute correlation (matching rate of multiple source attributes of the same entity). The coordinate transformation component supports automatic conversion of multiple reference systems such as WGS84 and the 2000 National Geodetic Coordinate System, as well as custom reference system configuration. The spatial index adopts a hybrid index structure of KD-Tree and R-tree to improve the retrieval efficiency of multiple types of features.

[0018] The spatial computing engine of the intelligent verification engine adopts the Spark distributed computing framework to achieve sharded parallel processing for tens of millions of vector data. The geometric operation includes six types of error detection: self-intersection, non-closure, overlap, gap, and dangling point. The Bayesian network model of the conflict arbitration engine is trained with historical verification data and has an accuracy of no less than 92% in judging contradictory results.

[0019] The cost-benefit model of the automatic repair module quantifies the repair solution from three dimensions: data accuracy retention rate, business logic compliance, and repair time. The automatic repair response time for simple errors is ≤500ms, the generation time for candidate solutions for complex errors is ≤3s, and the secondary verification pass rate must reach more than 98% to complete the repair process.

[0020] The traceability information of the full-process management module includes data source, verification time, repair operator, rule version, and error confidence. The weights of each dimension in the four-dimensional quality assessment model can be customized by the industry, and the quality score forms a trend comparison curve with historical data.

[0021] The federated learning framework of the rule optimization module adopts a horizontal federation model, which aggregates multi-node verification data while protecting data privacy. The reinforcement learning algorithm uses "improved verification accuracy + reduced false negative rate" as the reward function. The rule iteration cycle can be configured to 7-30 days. After each iteration, the rule base update log is synchronized to the blockchain.

[0022] System deployment and core component configuration This system is deployed using a cloud-edge-device collaborative B / S architecture to adapt to data processing needs in various scenarios: Cloud Node: Configured with a distributed computing cluster (including 3 nodes equipped with Intel Xeon Gold 6330 processors, 128GB of memory, and 20TB of storage), running the Spark distributed computing framework, PostgreSQL+PostGIS spatial database, and federated learning node service, responsible for parallel processing of large-scale data and rule optimization across scenarios; Edge nodes: Deploy industrial-grade edge computing gateways (equipped with Intel Core i7-12700 processors and 32GB of memory) to provide preprocessing services and local rule base caching for data access, thereby enabling fast format parsing and noise filtering of data sources and reducing cloud transmission pressure; Client-side: Supports multiple mainstream browsers (such as Chrome 110+, Edge 109+, and the domestic Kylin Browser), and enables smooth visualization of large-scale vector data through WebGL technology without the need to install special plugins.

[0023] Core algorithm component configuration: The conflict arbitration engine, which adopts a Bayesian network model, has been trained on 100,000 historical verification data and achieved a 93.2% accuracy rate in judging conflict results after 500 rounds of iteration; the reinforcement learning algorithm of the rule optimization module adopts the DQN (Deep Q Network) architecture, and the configured reward function weight is "60% increase in verification accuracy + 40% reduction in false negative rate"; the improved edge matching algorithm of the spatial computing engine is accelerated by GPU (using NVIDIA A100 graphics card), improving geometric calculation efficiency by 2.5 times.

[0024] Complete implementation process (data verification scenario for land and transportation integration) This implementation example aims to integrate and verify the "territorial spatial planning data" and "intelligent transportation road network data" of a provincial capital city. The processed data includes 800,000 polygonal features (land use types), 1.5 million linear features (expressways / main roads / secondary roads), and 3 million point features (traffic lights / bus stops). The data formats cover Shapefile (territorial data), GeoJSON (main road data), and GDB (transportation facility data). The specific implementation steps are as follows: Data access and feature enhancement preprocessing: After receiving multi-source data, edge nodes automatically parse the format and start the preprocessing process: The coordinate transformation component converts land data (2000 National Geodetic Coordinate System) and traffic data (WGS84 coordinate system) to the city-independent coordinate system; The hybrid indexing unit constructs a hybrid index of KD-Tree (point features) and R-tree (line / polygon features), improving retrieval efficiency by 40% compared to a single KD-Tree; The noise filtering unit removes 127 abnormal traffic light points (coordinates deviating from the road range by more than 50 meters) and 32 broken main road segments through "coordinate range verification + neighborhood density analysis"; The feature extraction unit simultaneously calculates data features, generates feature labels (average number of edges for polygon features: 18.6, average number of vertices for line features: 23.2, multi-source attribute matching rate for the same road entity: 89.3%), and uploads the preprocessed data and feature labels together to the cloud-based distributed database.

[0025] Rule pre-optimization and multi-dimensional configuration: The topology rule management module loads three levels of basic rules, and the basic rule library enables eight types of general rules, such as "surface-to-surface non-overlapping" and "line-to-line non-intersection (roads at the same level)". The semantic rule library loads land and transportation-specific rules, including "highway line elements should maintain a distance of ≥50 meters from permanent basic farmland surface elements", "bus stop elements must fall within 50 meters of urban main road or secondary road line elements", and "commercial ground elements must not contain traffic light point elements". At the same time, custom rule units can be configured through visual drag and drop with "surface element overlap area threshold ≤0.05 hectares" and "road line element suspension point length ≤2 meters".

[0026] The rule optimization module simultaneously initiated pre-optimization and utilized historical verification data (a total of 2.3 million error records) from the relevant city and three surrounding prefecture-level cities over the past three months. Through reinforcement learning algorithms, iterative optimization of the rule parameters was performed, adjusting the "road-farmland distance threshold" from a fixed 50 meters to a dynamic threshold of 50 meters in urban built-up areas and 80 meters in suburban areas. The "bus stop offset threshold" was further optimized to adapt to road classifications, resulting in a threshold of 60 meters for main roads and 40 meters for secondary roads. After review by experts in land and transportation, the optimized rules were updated to the rule base and stored on the blockchain.

[0027] Three-engine collaborative verification and conflict arbitration: The verification task is launched in the cloud, and the Spark cluster divides the data into 16 parallel tasks: The ontology reasoning engine loads the optimized semantic rules and uses the Pellet inference engine to detect two types of semantic conflicts: 15 instances of "commercial land contains traffic lights" and 8 instances of "bus stop offset is out of range". The spatial computing engine located geometric errors by improving the edge matching algorithm and hybrid indexing: it found 23 overlapping polygon features (including 7 overlapping land and transportation land boundaries), 19 self-intersecting features of main road lines, and 46 hanging points of road lines. The conflict arbitration engine intervened to handle three contradictory results, such as "a section of expressway should semantically be adjacent to the ecological protection red line surface element, but geometrically overlaps by 3.2 meters." Combining historical data (approval records for road expansion in the area in 2023) and a Bayesian network for probabilistic analysis (assessing a 91.5% probability that the conflict was caused by data update lag), it was ultimately determined to be a "geometric data lag error" and assigned a corresponding confidence level. The generated verification report includes the error type, location information, confidence level, and relevant evidence.

[0028] Cost-oriented tiered repair and secondary verification: Based on the error type and confidence level, the system initiates a tiered repair procedure. For automatic repair of simple errors: Automatic repair was performed on 19 road self-intersections (using the pruning method), 46 hanging points (using the extension connection method), and 8 small-area surface overlaps (using the buffer fusion method), with an average response time of approximately 320ms. Complex errors are addressed using cost-assisted decision-making: candidate solutions were set up for errors such as traffic lights on 15 commercial land sites and roads overlapping with ecological red lines on 3 sites. A cost-benefit model was used to quantify the errors from three dimensions: data accuracy (preserving approval document information), business logic (compliance with land spatial planning), and the time required for repair. In particular, for the overlap of roads with ecological red lines, Solution 1, "correcting road coordinates" (95% accuracy, 100% logical compliance, 2.1 seconds), received the highest score. The system displayed a preview of the repair effect, and the user performed the repair operation after confirmation.

[0029] The repaired data was automatically submitted for secondary verification, with a pass rate of 99.2%. The two errors that failed were legacy issues of "ambiguous ownership boundaries," which were marked by the system as "requiring manual on-site verification" and linked to relevant ownership files.

[0030] Full-process traceability and quality assessment: Utilizing blockchain technology, nodes record information throughout the entire lifecycle, including data source (e.g., Municipal Natural Resources Bureau / Municipal Transportation Bureau), verification time (e.g., 2025-10-28 09:15:32), repair operator (e.g., Mr. Li, land data reviewer), rule version (e.g., V2.3, after optimization), and error confidence level, forming an immutable traceability chain. A four-dimensional quality assessment model is constructed for scoring, with completeness at 96 points (error repair rate 98.7%), accuracy at 95 points (verification pass rate 99.2%), consistency at 93 points (multi-source data matching degree 92.1%), and timeliness at 94 points (time from data access to archiving 2.3 hours). The quality score and the results of the last three verifications form a trend curve, showing a gradual improvement in the "accuracy" dimension.

[0031] Finally, the generated verification report (supporting PDF / Excel format) and the repaired data are archived to the city's big data platform, and a task scheduling unit is configured to implement scheduled tasks such as "automatically triggering the next full verification at the end of each month" and "triggering incremental data verification every Wednesday".

[0032] Rule Iteration and Update: The rule optimization module aggregates the 230 error records and repair results from this verification, and combines them with verification data (after privacy anonymization) from three surrounding prefecture-level cities using a horizontal federated learning framework. The reinforcement learning algorithm completes the next round of rule iteration, further optimizing the "bus stop offset threshold" to "dynamic adjustment based on pedestrian flow (50-80 meters on main roads)." After the update log is synchronized to the blockchain, the rule base version is upgraded to V2.4.

[0033] To verify the actual effect of the newly added technical points of this invention, three sets of comparative tests were conducted, using "traditional dual-engine system (conflict-free arbitration and rule optimization)," "the complete system of this invention," and "a mainstream GIS platform." The test data was the aforementioned land and transportation fusion dataset, and the results are shown below: Test results show that the present invention effectively solves the problem of missed conflict judgment in traditional systems through the conflict arbitration engine, and the rule optimization module significantly improves cross-scenario adaptability. The overall verification efficiency and accuracy have formed a generational advantage over existing technologies, and fully meet the verification needs of complex scenarios such as land and transportation data integration.

[0034] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An automatic verification and management system for vector data topology, characterized in that, It includes a data access module, a topology rule management module, an intelligent verification engine, an automatic repair module, a full-process management module, a visual interaction module, and a rule optimization module. Each module interacts with other modules through a data bus. The data access module is used to receive multi-source vector data, complete format parsing, coordinate transformation, spatial index construction and noise filtering preprocessing. During the preprocessing process, the feature extraction unit mines the geometric complexity and attribute correlation of the vector data, and stores the preprocessed data and feature information into a distributed database. The topology rule management module constructs a three-level rule system of "basic rule library + semantic rule library + custom rules". The basic rule library contains 12 types of general topology constraint rules based on the Egenhofer nine intersection model. The semantic rule library combines domain ontology and SWRL rule language to define industry-specific constraints. Custom rule units support visual drag-and-drop configuration of rule parameters and logical associations. The intelligent verification engine adopts a three-engine architecture of "ontology reasoning + spatial computing + conflict arbitration". The ontology reasoning engine integrates the Pellet inference engine to realize semantic consistency verification. The spatial computing engine uses an improved edge matching algorithm combined with KD-Tree spatial index to realize geometric topology error detection. The conflict arbitration engine performs probabilistic analysis on the dual-engine verification results and historical data based on the Bayesian network model to solve the contradiction problem of "semantically valid but geometrically conflicting", and finally generates a verification report with confidence label. The automatic repair module performs hierarchical repair based on the error type and confidence level of the verification report. Simple errors are automatically repaired using pruning and buffer fusion methods, while complex errors generate multi-dimensional candidate solutions and quantify the repair impact through a cost-benefit model. After user confirmation, the repair operation is executed, and the repaired data needs to be verified a second time. The full-process management module records traceability information throughout the entire lifecycle using blockchain technology, constructs a four-dimensional quality assessment model of "completeness-accuracy-consistency-timeliness", and supports scheduled and triggered task scheduling and report archiving and export. The visualization and interaction module provides GIS map overlay display, error location, rule configuration panel and quality dashboard, and supports multi-role permission management based on RBAC model; The rule optimization module collects multi-scenario verification data through a federated learning framework and combines reinforcement learning algorithms to dynamically iteratively optimize the parameter thresholds and logical weights of the topology rules. The optimized rules are then manually reviewed and updated to the rule base.

2. The system according to claim 1, characterized in that, The feature extraction unit of the data access module generates data feature labels through geometric complexity calculation and attribute correlation analysis. The coordinate transformation component supports automatic conversion of multiple reference systems such as WGS84 and 2000 national geodetic coordinate system and custom reference system configuration. The spatial index adopts a hybrid index structure of KD-Tree and R-tree to improve the retrieval efficiency of multiple types of elements.

3. The system according to claim 1, characterized in that, The spatial computing engine of the intelligent verification engine adopts the Spark distributed computing framework to achieve sharded parallel processing for tens of millions of vector data. The geometric operation includes six types of error detection: self-intersection, non-closure, overlap, gap, and dangling point. The Bayesian network model of the conflict arbitration engine is trained with historical verification data and has an accuracy rate of no less than 92% in judging contradictory results.

4. The system according to claim 1, characterized in that, The cost-benefit model of the automatic repair module quantifies the repair solution from three dimensions: data accuracy retention rate, business logic compliance, and repair time. The automatic repair response time for simple errors is ≤500ms, the generation time for candidate solutions for complex errors is ≤3s, and the secondary verification pass rate must reach 98% or higher to complete the repair process.

5. The system according to claim 1, characterized in that, The traceability information of the full-process management module includes data source, verification time, repair operator, rule version, and error confidence. The weights of each dimension in the four-dimensional quality assessment model support industry customization, and the quality score forms a trend comparison curve with historical data.

6. The system according to claim 1, characterized in that, The federated learning framework of the rule optimization module adopts a horizontal federation mode, which aggregates multi-node verification data while protecting data privacy. The reinforcement learning algorithm uses "improved verification accuracy + reduced false negative rate" as the reward function. The rule iteration cycle can be configured to 7-30 days. After each iteration, the rule base update log is synchronized to the blockchain.

7. A vector data topology verification method based on the system according to any one of claims 1-6, characterized in that, Includes the following steps: S1: The data access module receives multi-source vector data, completes preprocessing and feature extraction, and stores the data and feature labels into a distributed database; S2: The topology rule management module loads the three-level rules, and the rule optimization module calls historical data to pre-optimize the rules; S3: The intelligent verification engine initiates a three-engine collaborative verification process, the conflict arbitration engine handles contradictory results, and generates a verification report with confidence level. S4: The automatic repair module performs hierarchical repair based on the verification report. For complex errors, a cost-benefit model is used to assist in decision-making. After repair, the data is verified a second time. S5: The full-process management module records traceability information and generates a four-dimensional quality score, and archives data and reports; S6: The rule optimization module aggregates the verification data, completes rule iteration through reinforcement learning, and updates the rule base.

Images