Traffic Sign Recognition System
Maqsood Ahmed, Raja Hashim Ali, Nisar Ali
Abstract
Traffic Sign Recognition Systems (TSRS) are instrumental in improving road safety by assisting drivers and supporting autonomous vehicles in real-time identification of road regulations. These systems have advanced significantly in recent years, with deep learning approaches achieving promising results. However, despite these developments, existing models face challenges in adapting to the diverse conditions of real-world environments, including variations in lighting, weather, and sign appearance. Addressing this limitation, our study focuses on enhancing the robustness and accuracy of TSRS for practical deployment. In response to this gap, we introduce a novel deep learning model optimized for traffic sign recognition across various conditions. Our approach utilizes a convolutional neural network (CNN) architecture, which was trained on the German Traffic Sign Recognition (GTSR) dataset. We improved the model’s adaptability to new and unseen data while achieving a high accuracy rate of 97.99% by applying techniques like data augmentation and transfer learning. Methodologically, the model workflow includes extensive preprocessing, hyperparameter tuning, and real-time inference evaluations, ensuring suitability for deployment in autonomous driving systems. Our study contributes by providing a scalable, high-accuracy TSRS model that can reliably identify traffic signs in complex environments, supporting both autonomous and assisted driving technologies. The model’s performance paves the way for safer and more efficient road transport, while our results highlight the potential for further interdisciplinary research to expand TSRS capabilities.