Comparative Study and Analysis of Different Target Detection Algorithms in Traffic Sign Detection

DOI: https://doi.org/10.62517/jes.202302213

Author(s)

Chunli Wang, Long Xiangyu, Chu Zhongmin, Pan Mingfang

Affiliation(s)

School of Tourism Data, Guilin Tourism University, Guilin, Guangxi, China

Abstract

The advancement of artificial intelligence in transportation has led to a burgeoning interest in the research of automatic identification technologies, particularly in the realm of traffic signs. It is an important pioneer technology of unmanned driving technology and has great theoretical value and application prospect. However, traffic sign detection is faced with the influence of complex weather factors such as rain, snow and fog, as well as the problem that the target is partially blocked and the size of the target is very small. Hence, selecting a target detection algorithm capable of swiftly and precisely identifying traffic sign categories is imperative. This paper compared various target detection algorithms, trained and tested YOLO v3, YOLO v4, SSD and other algorithms using the same traffic sign data set (30 classes), and finally concluded that the YOLO v4 network had the best effect, with a mAP value of 83.28% and a convergence interval of total loss between 3.5 and 4.

Keywords

Traffic Sign Recognition; YOLO; SSD; Algorithm Contrast

References

[1]Jiang Gangyi,Zheng Yi. Automatic traffic sign recognition based on mathematical morphology. Journal of Shantou University(Natural Science Edition),1998,1998(013):90-96. [2]Wu Haomin.Research and Implementation of Road Traffic Sign Recognition Based on Driving Video.Nanjing University of Posts and Telecommunications for the Degree of Master of Engineering,2020. [3]Yang Yanfei,CaoYang.Improved glass insulator detection in Yolov3 drone shot.Computer Engineering and Applications,2021,1-11. [4]Zheng Z,Wang P,Liu W,et al.Distance-IoU Loss: Faster and Better Learning for Bounding Box Regression. arXiv, 2019. [5]Li Zhou,Huang Miaohua.Real-time vehicle detection based on YOLO_v2 model.Chinese Journal of Mechanical Engineering,2018,29(15):1869-1874. [6]Redmon J,Divvala S,Girshick R and Farhadi A. You Only Look Once: Unified, Real-Time Object Detection.2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR), Las Vegas, NV, United States, 2016: 779-788. [7]Garcia C,Delakis M.Convolutional face finder:A nerual architecture for fast and robust face detection.IEEE Transactions on Pattern Analysis and Machine Intelligence,2004, 26(11):1408-1423.