Classification of traffic signs in real-time on a multi-core processor

This paper presents a neural network approach to classify traffic signs based on greyscale images. The developed system runs on a multi-core processor. The optimization of the neural network concerning fix-point arithmetic and memory consumption results in real-time implementation without the requirement of an external memory (low system costs). A parallelization of the processing scheme allows a high utilization of the multi-core processor. The neural network proposed in this paper is trained with computer generated samples of traffic signs. These patterns cover most possible distortions and main environment situations.