Vietnamese License Plate Recognition Using CNN

Abstract

This work shows an approach for constructing a system for detecting and recognizing license plates (LPDR). In the first stage, the poly region of the license plate’s word line(s) can be detected by Character-Region Awareness for Text Detection (CRAFT). Especially, the text line of one-line license plate and two lines of the multi-line license plate can be detected effectively. Second, the Mobinet architecture will be used to extract features from each region offered by CRAFT as a plate number region. Finally, these characteristics will be transmitted into’Bi's long-term memory (Bi-LSTM) architecture with Connectionist Temporal Classification (CTC) to predict output text in each input region. By applying this solution, the problem relating to multi-line license plates can be appropriately handled.