Convolutional Neural Network Based Traffic Sign Recognition System

As an important part of intelligent transport system (ITS), road traffic sign recognition (TSR) system gathers information from real-time video frames and recognizes the content of the traffic signs. The analysis and process of ITS mainly depends on the speed and accuracy of TSR system, which are the key factors to improve driver safety. We proposed a method for real time traffic sign recognition based on convolutional neural network (CNN). The training database was established by field sample collection, with which the neural network model was trained. Stochastic gradient descent (SGD) optimizer is utilized during training to improve the learning efficiency. The test results show that the proposed method achieves good performance in speed, accuracy and robustness for real time TSR.