Traffic sign representation using sparse-representations

Abstract

Automatic Traffic Sign Recognition has gained significant impetus among the research community in recent times. Increasing demands in the arenas of Autonomous Vehicle Navigation and Driver Assistance Systems is making this field of research more attractive. In this paper, we developed a technique which uses Sparse Representation based Classification coupled with Boundary Discriminative Factor (BDF) for recognizing traffic signs. The performance of this system is compared with one of the existing classifiers, Convolutional Neural Networks (CNNs) which has been employed in many real-time systems. This method also helps in reducing the enormous training time required for CNNs.