Title: FastVGG network and its application in automatic identification of traffic signs

Authors: Kaige Gui; Shangbing Gao; Xinyuan Li; Zhonghe Lu

Addresses: College of Computer and Software Engineering, Huaiyin Institute of Technology, Huaian, 223001, China; Jiangsu Mobile Internet of Things Technology Engineering Laboratory, Huaian 223001, China ' College of Computer and Software Engineering, Huaiyin Institute of Technology, Huaian, 223001, China; Jiangsu Mobile Internet of Things Technology Engineering Laboratory, Huaian 223001, China ' College of Computer and Software Engineering, Huaiyin Institute of Technology, Huaian, 223001, China; Jiangsu Mobile Internet of Things Technology Engineering Laboratory, Huaian 223001, China ' College of Computer and Software Engineering, Huaiyin Institute of Technology, Huaian, 223001, China; Jiangsu Mobile Internet of Things Technology Engineering Laboratory, Huaian 223001, China

Abstract: Traffic sign recognition is an important part of driver assistance systems. Because the types of traffic signs are complex and diverse, they are difficult to identify. Traditional recognition methods require manual extraction of features, which is difficult and inaccurate. This paper proposes a FastVGG network based on VGG neural network to extract the features of the target image to realise the recognition of traffic signs under different angles and illumination. In the method of this paper, the connection layer parameters and the number of network layers are moderately reduced, the merging step is increased, and the recognition speed is improved. When the parameter value is less than zero, the leaky ReLU is used to replace the activation function to solve the problem of neuron death. The experimental results of the German traffic sign recognition dataset (GTSRD) show that the algorithm can achieve accurate classification while increasing the speed.

Keywords: traffic sign recognition; driving assistant system; deep learning; VGG Net.

DOI: 10.1504/IJICT.2022.120636

International Journal of Information and Communication Technology, 2022 Vol.20 No.2, pp.192 - 203

Received: 29 Apr 2020
Accepted: 06 Aug 2020
Published online: 31 Jan 2022 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article