Character Time-series Matching For Robust License Plate Recognition

Automatic License Plate Recognition (ALPR) is becoming a popular study areaand is applied in many fields such as transportation or smart city. However,there are still several limitations when applying many current methods topractical problems due to the variation in real-world situations such as lightchanges, unclear License Plate (LP) characters, and image quality. Almostrecent ALPR algorithms process on a single frame, which reduces accuracy incase of worse image quality. This paper presents methods to improve licenseplate recognition accuracy by tracking the license plate in multiple frames.First, the Adaptive License Plate Rotation algorithm is applied to correctlyalign the detected license plate. Second, we propose a method called CharacterTime-series Matching to recognize license plate characters from manyconsequence frames. The proposed method archives high performance in theUFPR-ALPR dataset which is \boldmath$96.7\%$ accuracy in real-time on RTX A5000GPU card. We also deploy the algorithm for the Vietnamese ALPR system. Theaccuracy for license plate detection and character recognition are 0.881 and0.979 $mAP^{test}[email protected] respectively. The source code is available athttps://github.com/chequanghuy/Character-Time-series-Matching.git