Traffic Sign Recognition Using Convolutional Neural Network and Skipped Layer Architecture

Abstract

This paper proposes an efficient deep learning approach for the Traffic sign recognition system (TSR). TSR plays an essential role in the intelligent transportation system (ITS). TSR can be utilised for driver assistance and, eventually, driverless vehicles to reduce accidents. Drivers usually concentrate on the road when driving; however, most traffic signs are located on the side of the road, leading to an accident. TSR enables users to observe traffic sign information without having to divert their attention away from the road. For the past four decades, TSR has been a hot topic of research. Because of the broad background, clutter, varying levels of illumination, variable sizes of traffic signs, and fluctuating weather conditions, TSR is an important but challenging procedure in ITS. Many efforts have been made to find answers to the major issues that they face. The objective of this study addresses the road traffic sign recognition using skipped convolutional layer architecture and neural network. This study differs from earlier studies in that it uses skipped layer connections and extracts information in two ways. The features are then concatenated and fed into a fully connected neural network. Finally, SoftMax is used to predict the labels of the traffic sign. The German Traffic Sign Recognition Benchmark (GTSRB) dataset was used to train our proposed Convolutional Neural Network (CNN) model. The model was tested on 43 different traffic sign classes using 12630 images from the GTSRB dataset. The same dataset had a classification accuracy of 99.4%, whereas the existing lenet-based model had an accuracy of 97.1%. Rather than a more intricate model, the effort produced a basic yet effective model for recognising traffic signs.