A new designed descriptor for road sign recognition

Abstract

Road sign recognition (RSR) systems are one of the main tasks of intelligent transportation systems (ITS). These systems employ vehicle mounted cameras to identify traffic signs while driving on the road. Their primary function is to inform the driver of recent traffic signs that may have been missed due to distraction or inattentiveness. In this work, a new method for road sign detection and recognition is proposed. The proposed approach is divided into three stages: first, a color segmentation method is used to extract regions of interest (ROIs). Then, we refer to polygonal approximation technique to detect triangular, rectangular, and circular shapes. The last stage aims at recognizing the detected signs using a new designed feature and SVM classifier. The proposed approach was tested on two publicly available datasets, and the results obtained are satisfying compared to the state-of-the-art methods.