An Intelligent Vehicle Recognition System for Access Control in Algeria

Abstract

Vehicle recognition systems are essential tools in modern intelligent transportation and security infrastructures, enabling automated access control, traffic monitoring, and law en forcement. In this thesis, we propose an intelligent vehicle recognition system specifically designed for access control in Algeria. The system integrates several deep learning-based modules to identify and analyze key vehicle attributes, including license plate detection and recognition, vehicle orientation, color, brand, and model. The system handles both image and video inputs. To achieve this, we adopt the YOLOv8 model for multiple tasks, including vehicle detection, license plate localization, orientation estimation, and logo de tection. The well-known OCR technique is also applied during the character recognition phase, while Convolutional Neural Networks are employed for vehicle color classification and logo recognition. Finally, MobileNet is used for vehicle model identification. Different datasets were applied, with a specific dataset used for each phase. The proposed multi stage architecture achieves high accuracy, indicating strong performance and efficiency, which leads to a robust and intelligent vehicle recognition pipeline, offering a comprehen sive solution that supports the core components of secure and automated access control systems.