Abstract
Automatic License Plate (LP) detection and recognition algorithms have become a necessity for intelligent transportation systems due to their efficiency in multiple applications such as parking control and traffic management. Vehicle LP detection and recognition is the process of identifying and locating the LP from the vehicle and then extracting and recognizing the characters from this plate. Several academic studies have addressed the problem of LP detection and recognition and have proposed implementations with different performance indicators’ values. However, many of the current studies’ solutions are still not robust and efficient in complex real-world situations. In this paper, an automatic LP detection and recognition approach is proposed for the context of Saudi Arabia using Deep Learning (DL) techniques. The core of the proposed approach is to develop a sequence of Faster Region-based Convolutional Neural Networks (Faster-RCNN) and Convolutional Neural Networks (CNN). The Faster-RCNN model is used for LP detection, whereas CNN is applied for characters’ recognition from LPs. The obtained experimental results prove the robustness and effectiveness of the proposed approach. We obtained a precision of 92% for LP detection and an accuracy of 98% for LP recognition.