Traffic sign detection by template matching based on multi-level chain code histogram

Abstract

This paper proposes a real-time system for traffic signs detection, which features of template matching based on a new feature expression for geometric shapes, namely, multi-level chain code histogram (MCCH). For all of the different shapes associated with Chinese traffic signs, e.g., circle, triangle, inverted triangle and octagon, MCCH is a robust feature expression with remarkable low computational cost, which is particularly important for real-time applications. The proposed system consists of three stages: 1) segmentation based on color; 2) MCCH extraction; 3) template matching. Extensive experiments were conducted using different datasets, demonstrating outstanding performance with regard to high processing speed and accuracy. The system robustness to rotation, scale, and illumination has also been illustrated.