Research on Classification of Traffic Signs Based on Convolutional Neural Network

In the intelligent transportation system, traffic signs are hindered by external and internal factors such as light, weather and shelter, which is easy to lead to identification errors. In view of this problem, this paper selects the identification of traffic signs based on LeNet-5 network model. In order to reduce the burden of convolutional neural network and improve the resolution, this paper collects traffic sign images from the actual environment, and first performs a series of preprocessing operations for GTSRB data set, including image gray, size normalization and image enhancement. The preprocessed GTSRB data set helps identify subsequent convolutional neural network with LeNet-5 classical model. The experimental results show that the algorithm has high recognition rate and great advantages in traffic sign recognition.