The Research on Traffic Sign Recognition Based on Deep Learning

With the further and faster urbanization, here come the advent and development of intelligent public transportation system. The identification of the traffic signs, as a key component of the intelligent transit system, has a prospect of widespread use in self-driving and driver assistance system. The paper focuses on the classification and identification of traffic signs based on the researches on traffic signs home and abroad which also include the status quo, technological problems and development tendency. Combined with the significant features of the shape and internal structure of the traffic signs, three new feature-based algorithms on road signs identification have been implemented, including image pretreatment, feature extraction and classifier. The feature extraction work of traffic signs is conducted with the algorithms using an integration of Deep Boltzmann Machines and Canonical Correlation Analysis. Compared with the algorithms based on HOG and LBP feature extraction, the DBM-CCA has higher accuracy.