A New Fusion-Based Approach for License Plate Recognition: An Application to the Algerian Context

Automatic License Plate Recognition (ALPR) systems are widely used in a large field of applications such as surveillance, security monitoring, and traffic enforcement. Therefore, many efforts have been devoted to the improvement of the recognition accuracy of these systems. In this paper, we propose a fusion-based approach to handle Algerian license plates identification. The main purpose of our approach is to improve the recognition rate by using data fusion techniques. Two different methods have been used for plate characters segmentation, namely: deep learning YOLO object finder and edge detection. Initial results show the effectiveness of the first method compared to the latter. In addition, several classifiers are used for the plate character recognition stage. These classifiers have been combined using data fusion techniques, namely: evidence theory and majority voting method. To evaluate our proposals, many experiments have been conducted on a dataset of Algerian license plate images gathered by ourselves from real world conditions. The obtained results show that the proposed fusion-based approach reaches good performances compared to the baselines.