Hierarchical Traffic Sign Recognition

Abstract

Traffic Sign Recognition (TSR) is very important for driverless systems and driver assistance systems. Because of the large number of the traffic sign classes and the unbalanced training data, we propose a hierarchical recognition method for traffic sign recognition. A classification tree is constructed, where the non-leaf node is constructed based on shape classification with aggregated channel features and a leaf node is constructed based on random forest classifiers with histogram of gradient for multi-class traffic sign recognition in the non-leaf node. The proposed method can overcome the inefficiency of flat classification scheme and imbalance of training data. Extensive experiments are done on three famous traffic sign datasets: the German Traffic Sign Recognition Benchmark (GTSRB), Swedish Traffic Signs Dataset (STSD), and the 2015 Traffic Sign Recognition Competition Dataset. The experimental results demonstrate the efficiency and effectiveness of our methods.