CN-122019679-A - Efficient map element quality inspection AI system and implementation method

CN122019679ACN 122019679 ACN122019679 ACN 122019679ACN-122019679-A

Abstract

The invention provides an efficient map element quality inspection AI system and an implementation method, which relate to the field of map element quality inspection systems and comprise a multi-source data access module, a multi-source data acquisition module and a multi-source data acquisition module, wherein the multi-source data access module is used for importing multi-type map element data and unifying a space reference system; the AI quality inspection model training module comprises a space topological feature extraction network and a quality inspection rule self-learning sub-model for generating a dynamic rule base, a multi-element intelligent verification module, an integrated topological verification unit, an attribute verification unit and an association relation verification unit, a visual report and correction module, a system management module and a user authority management and model iteration maintenance module, wherein the integrated topological verification unit, the attribute verification unit and the association relation verification unit are used for joint verification of multi-type map elements, the visual report and correction module is used for generating a structured quality inspection report and supporting one-key correction. 8 common error types in map drawing can be automatically identified, and efficient, accurate and intelligent map quality inspection is realized by combining multi-mode data with deep learning and knowledge reasoning.

Inventors

MA LI
HAN LINHUA
LIU XIN
DONG DECHENG
SHI DONG
LI XIAOFANG

Assignees

中国人民解放军61206部队

Dates

Publication Date: 20260512
Application Date: 20260129

Claims (10)

1. An efficient map element quality inspection AI system, comprising: the multi-source data access module is used for importing multi-type map element data and unifying a space reference system; the AI quality inspection model training module comprises a space topological feature extraction network and a quality inspection rule self-learning sub-model and is used for generating a dynamic rule base; The multi-element intelligent verification module is integrated with a topology verification unit, an attribute verification unit and an association relation verification unit and is used for joint verification of multi-type map elements; The visual report and rectification module is used for generating a structured quality inspection report and supporting one-key rectification; And the system management module is used for user authority management and model iterative maintenance.
2. The AI system for quality inspection of map elements according to claim 1, wherein the multi-source data access module supports the introduction of main stream GIS vector data formats such as SHP, GDB, geoJSON, PDF and satellite grid base map data, and a built-in space reference system automatic conversion unit can uniformly convert map elements of different coordinate systems into a target coordinate system, and the target coordinate system comprises CGCS2000 and WGS84.
3. The efficient map element quality inspection AI system of claim 1, wherein the spatial topology feature extraction network is based on an improved PointNet ++ algorithm, introduces GIS spatial neighborhood relation weight factors, performs high-dimensional feature coding on node coordinates, geometric forms and adjacent element association relations of map elements, and automatically generates and iteratively updates a dynamic verification rule base adapting to multiple scenes by adopting a reinforcement learning framework, historical quality inspection marking data as training samples and compliance indexes of different map publishing/production standards as rewarding functions by a self-learning sub-model of quality inspection rules.
4. The efficient map element quality inspection AI system as set forth in claim 1, wherein the topology verification unit is configured with a node micro-fracture recognition threshold, an element overlapping area threshold, a self-intersection judgment algorithm, and is capable of automatically recognizing hidden topology errors such as node fracture, geometric overlapping, self-intersection and the like of the map element, the attribute verification unit is capable of verifying integrity of necessary filling fields of the map element, compliance of field value fields and consistency of attribute association based on a dynamic rule base, the attribute association comprises matching relation of road grade and speed limit value and corresponding relation of river-crossing bridge and water system grade, the association relation verification unit fuses a space overlapping analysis algorithm and an attribute matching algorithm, and joint verification of space position rationality and attribute association can be carried out for multiple element combinations such as road-water system, road-traffic accessory, water system-administrative division and the like.
5. The system of claim 1, wherein the visual report and correction module supports generation PDF, excel, GIS of three types of quality control reports including spatial location labels, error type classifications, correction suggestions of error elements, and configuration of a one-key correction unit for automatically correcting common errors such as node micro-breaks, field format non-norms, etc.
6. The system of claim 1, wherein the system management module comprises a task log recording unit and a model iteration triggering unit, the task log recording unit stores data source information, a verification rule version and an error correction record of each quality inspection, and the model iteration triggering unit can automatically start an iteration training process of the AI quality inspection model based on manual correction feedback data.
7. The method for implementing an efficient map element quality inspection AI system of claims 1-6, comprising the steps of: s1, preprocessing data, namely importing map element data through a multi-source data access module to finish data format analysis, space reference system unified conversion, repeated element elimination and ineffective geometric cleaning; S2, training an AI quality inspection model, namely constructing a training data set containing samples of topology errors, attribute errors and association relation errors, extracting spatial topology features by adopting an improved PointNet ++ algorithm, and generating a dynamic rule base by self-learning a sub-model of a reinforced learning training quality inspection rule; S3, inputting the preprocessed map element data into a pre-training model, sequentially executing topology verification, attribute verification and association relation verification, integrating verification results and marking the unique identification and the spatial position of the error element; S4, report generation and correction, namely outputting a structural quality inspection report through a visual report and correction module, supporting one-key correction of general errors, and inputting corrected data into a system again for rechecking; S5, model iterative optimization, namely inputting the rechecking result and the manual correction feedback data into an AI quality inspection model training module, and continuously optimizing model parameters and a dynamic rule base.
8. The method according to claim 7, wherein the constructing the training data set in step S2 includes collecting map quality inspection data of different scales and different application scenes, and performing three-level labeling on the error samples, where the three-level labeling includes error type, error severity, and correction priority.
9. The method for implementing the efficient map element quality inspection AI system as claimed in claim 7, wherein the specific process of the association verification in step S3 is that firstly, the spatial position relationship of two types of elements is judged through a spatial superposition analysis algorithm, then, the consistency of association fields among the elements is checked through an attribute matching algorithm, and finally, whether the association is in compliance is judged by comparing with a dynamic rule base.
10. The method for implementing the efficient map element quality inspection AI system according to claim 7, wherein in the step S5, the model iterative optimization adopts an incremental training mode, and only the feature dimensions corresponding to the newly added feedback data are updated in parameters, so that the model is not required to be retrained in a full quantity, and the iteration period is greatly shortened.

Description

Efficient map element quality inspection AI system and implementation method Technical Field The invention relates to the technical field of map element quality inspection systems, in particular to an efficient map element quality inspection AI system and an implementation method. Background In the process of map making and updating, the quality of map element data is important. The traditional map element quality inspection method mainly relies on manual inspection, and the method is low in efficiency, is easily affected by human factors, and is difficult to ensure consistency and accuracy of quality inspection results. With the rapid development of the GIS technology, the data volume of map elements is rapidly increased, the data types are increasingly complex, and the traditional quality inspection method cannot meet the quality inspection requirements of large scale, high efficiency and high precision. At present, although some automatic quality inspection tools exist, most of the automatic quality inspection tools are single in function, only can be used for simply checking specific types of map elements or specific error types, and the automatic quality inspection tools lack the joint checking capability of the map elements of multiple types, are fixed in quality inspection rules, and are difficult to adapt to the changes of different application scenes and map publishing/production standards. Therefore, there is a need for an efficient map element quality inspection AI system. Disclosure of Invention In order to solve the problems in the background technology, the invention provides an efficient map element quality inspection AI system and an implementation method, which can automatically identify 8 common error types in map drawing, and realize efficient, accurate and intelligent map quality inspection by combining multi-mode data with deep learning and knowledge reasoning. The invention provides an efficient map element quality inspection AI system, which specifically comprises: the multi-source data access module is used for importing multi-type map element data and unifying a space reference system; the AI quality inspection model training module comprises a space topological feature extraction network and a quality inspection rule self-learning sub-model and is used for generating a dynamic rule base; The multi-element intelligent verification module is integrated with a topology verification unit, an attribute verification unit and an association relation verification unit and is used for joint verification of multi-type map elements; The visual report and rectification module is used for generating a structured quality inspection report and supporting one-key rectification; And the system management module is used for user authority management and model iterative maintenance. Furthermore, the multi-source data access module supports the introduction of the main stream GIS vector data format such as SHP, GDB, geoJSON, PDF and the satellite grid base map data, and a space reference system automatic conversion unit is built in the multi-source data access module, so that map elements of different coordinate systems can be uniformly converted into a target coordinate system, and the target coordinate system comprises CGCS2000 and WGS84. Furthermore, the spatial topological feature extraction network introduces a GIS spatial neighborhood relation weight factor based on an improved PointNet ++ algorithm, performs high-dimensional feature coding on node coordinates, geometric forms and adjacent element association relations of map elements, and automatically generates and iteratively updates a dynamic verification rule base adapting to multiple scenes by adopting a reinforcement learning framework, taking historical quality inspection marking data as training samples and compliance indexes of different map publishing/production standards as rewarding functions by a quality inspection rule self-learning sub-model. Further, the topology verification unit is configured with a node micro-fracture recognition threshold, an element overlapping area threshold and a self-intersection judgment algorithm, can automatically recognize hidden topology errors such as node fracture, geometric overlapping, self-intersection and the like of map elements, and can check the integrity of necessary filling fields of the map elements, the compliance of field value fields and the consistency of attribute association based on a dynamic rule base, wherein the attribute association comprises a matching relationship between a road grade and a speed limit value and a corresponding relationship between a river-crossing bridge and a water system grade, and the association relationship verification unit fuses a space overlapping analysis algorithm and an attribute matching algorithm and can develop joint verification of space position rationality and attribute association aiming at various element combinations such as a road-water