Vehicle license plate detection using morphological operations and deep learning

Abstract

Vehicle License Plate Detection (VLPD) is the most critical stage of any vehicle License Plate Recognition (LPR) system because it has a direct impact on its robustness and accuracy. As a result, VLPD remains a difficult task because vehicle license plates (VLP) vary in size, axes, orientation, and may be occluded or have their locations changed. In this paper, we present our framework for an image-based VLPD system based on morphological information and deep learning. To address this issue, we created a new "YellowLP" dataset with 1050 images of unique and different rear VLP numbers. Pecision, recall, and overall accuracy of the morphological results are 98.65%, 97.90%, and 96.61%, respectively, with a detection rate of 97.90%. Deep learning increases the recall and overall accuracy of the proposed approach to 100% and 98.65%, respectively. As an outcome, the proposed method produced acceptable results.