Traffic signs shape recognition based on contour descriptor analysis

Abstract

Methods for automated traffic sign detection and recognition represents one of many structural pillars and basic building blocks in autonomous and self-aware transportation systems. Computationally efficient and recognition successful methods have great importance in accomplishments of such tasks. This paper presents a novel technique for traffic signs shape recognition that is based on analysis of the signs shape contour descriptors. The primary task of the presented method is proper recognition and classification of signs by their shapes into three basic categories; circle, triangle and square/rectangle one. The method utilizes a shape contour description in the manner of absolute angular tracing characteristics of a sign shape. Contour descriptors of different sign shapes in various imaging condition are analyzed, results are shown and evaluated. The presented method successfully distinct traffic signs by their shapes and proves the method usability.