High performance traffic sign recognition based on sparse representation and SVM classification

The high variability of sign appearance in uncontrolled environments has made the detection and classification of road signs a challenging problem in computer vision. In this paper, an occlusion-robust traffic sign recognition method is proposed. To achieve occlusion-robust detection, we design a cascaded tree detector based on the MN-LBP features and a cascaded tree. For occlusion-robust traffic sign classification, the occlusion-robust dictionaries for sparse representation of multiclass traffic signs are designed. Then, the results of sparse representation are classified with SVM method. The classification results of SVM are more robust than that of the sparse representation classification (SRC) which directly uses judgment. The experiments on test set show that the proposed method is more robust and accurate to detect signs with partial occlusion than the methods based on SVM or SRC.