A Novel Translation, Rotation, and Scale-Invariant Shape Description Method for Real-Time Speed-Limit Sign Recognition

Speed-limit sign (SLS) recognition is an important function to realize automatic driving assistance systems (ADAS). This paper presents a novel design of an image-based SLS recognition algorithm, which can efficiently detect and recognize SLS in real-time. To improve the robustness of the proposed SLS algorithm, this paper also proposes a new shape description method to describe the detected SLS using centroid-to-contour (CtC) distances of the sign content. The proposed CtC descriptor is invariant to translation, rotation, and scale variations of the SLS in the image. This advantage increases the recognition rate of a linear support vector machine classifier. The proposed SLS recognition method had been implemented and tested on an ARM-based embedded platform. Experimental results validate the SLS recognition accuracy and real-time performance of the proposed method.