Traffic Sign Classification Based on Deep Learning of Image Invariant Feature

Traffic signs captured in natural scene have various deformation and manual design method that used to extract invariant feature needs processing techniques. To address these problems, this paper proposed an automatic learning method for the traffic sign classification based on invariant feature extraction. Firstly, the feature map matrix of each stage is automatically learned by deep learning framework with slow feature analysis; Secondly, the first and the second stage features of traffic sign image are extracted and combined as the final traffic sign feature; Finally, traffic signs are classified by support vector machine(SVM). Experimental results show that the proposed method has good generalization performance, which can be effectively used in traffic sign classification, and the feature learned by this method is invariant to small translation and rotation.