A New Traffic Sign Recognition System with IFRS Detector and MP-SVM Classifier

The design of traffic sign recognition (TSR) system, one important subsystem of Advanced Driver Assistance System (ADAS), has been a challenge practical problem for many years due to the complex issues like road environments, lighting conditions, occlusion, and so on. In this paper, we introduce a new TSR system, whose effectiveness has been tested through extensive experiments. The established TSR system mainly consists of two parts, i.e. traffic sign detector and traffic sign classifier. In this system, the traffic sign detection is implemented with a new method based on improved fast radial symmetry detector, for detecting a class of circular prohibitive traffic signs efficiently and robustly. The traffic sign classification is accomplished through moments-based pictogram support vector machine (MP-SVM) classifier. Two kinds of features, Zernike Moments and Pseudo-Zernike Moments, are used to represent the pictogram, which will be fed to SVM for training and testing. Experiment results have validified the robust detection effects and high classification accuracy.