Real-time detection of traffic signs on a multi-core processor

Abstract

This paper presents the implementation of a traffic sign detection system on a multi-core processor. The approach, based on Viola-Jones, allows a fast, robust and reliable detection on grayscale images. The use of computer-graphics generated images supports an automatic process of the required generation of training data and a simple extension or adjustment to further traffic signs, different camera types and perspectives etc. The implementation fulfills the real-time constrain by consuming only 62.5% of the available computational power. An optimization of the memory consumption avoids the need of external memory.