Intelligent Text-Matching System for Compliance Verification of Commercial Vehicles in Active and Passive Safety Research
TL;DR
This system framework provides an efficient and interpretable solution for compliance verification, verifying the feasibility of lightweight AI in vertical fields and can also provide key data support for active and passive safety research such as the prevention of pedal misoperation in electric commercial vehicles.
Abstract
The compliance verification of operating vehicles, also known as the verification of compliant vehicles, is a crucial link in ensuring road transportation safety. However, the traditional manual interpretation of the complex remarks in the announcement of compliant vehicle models suffers from low efficiency and is prone to errors. Therefore, this study proposes a multi-level intelligent text matching (MLITM) system framework to automatically determine whether text fragments comply with and explain the differences between the actual vehicles and the parameters in the announcement during the verification process. Methodologically, MLITM adopts a three-level collaborative strategy: rapid filtering through full-text matching; detection of high-similarity segments through the longest common substring matching; and recall of scattered keywords through adaptive sliding window matching. Experiments show that the matching accuracy of MLITM on the simulation dataset is 103.4% higher than that of the regular expression method, with an average processing time of only 26 microseconds and an efficiency improvement of 26 times; among all methods compared in this study, MLITM achieves the best recognition accuracy and recall for various typical samples. Furthermore, by encapsulating the core MLITM algorithm as a microservice module and integrating it into an actual vehicle compliance review system, this study verifies that the proposed method can effectively assist auditors in rapidly locating supporting textual evidence. The results demonstrate its significant potential for improving review efficiency and enhancing the consistency of review outcomes.This system framework provides an efficient and interpretable solution for compliance verification, verifying the feasibility of lightweight AI in vertical fields. Additionally, based on the structured vehicle parameter database generated by this study, it can also provide key data support for active and passive safety research such as the prevention of pedal misoperation in electric commercial vehicles.
