first_page
Download PDF
settings
Order Article Reprints
Open AccessArticle
Deep Learning-Based Real-Time Traffic Sign Recognition System for Urban Environments
by
Chang-il Kim
,
Jinuk Park
,
Yongju Park
,
Woojin Jung
and
Yong-seok Lim
*
Korea Electronics Technology Institute, Seongnam 13509, Republic of Korea
*
Author to whom correspondence should be addressed.
Infrastructures 2023, 8(2), 20; https://doi.org/10.3390/infrastructures8020020
Submission received: 18 November 2022 / Revised: 14 January 2023 / Accepted: 26 January 2023 / Published: 31 January 2023
Downloadkeyboard_arrow_down
Browse Figures
Versions Notes
Abstract
A traffic sign recognition system is crucial for safely operating an autonomous driving car and efficiently managing road facilities. Recent studies on traffic sign recognition tasks show significant advances in terms of accuracy on several benchmarks. However, they lack performance evaluation in driving cars in diverse road environments. In this study, we develop a traffic sign recognition framework for a vehicle to evaluate and compare deep learning-based object detection and tracking models for practical validation. We collect a large-scale highway image set using a camera-installed vehicle for training models, and evaluate the model inference during a test drive in terms of accuracy and processing time. In addition, we propose a novel categorization method for urban road scenes with possible scenarios. The experimental results show that the YOLOv5 detector and strongSORT tracking model result in better performance than other models in terms of accuracy and processing time. Furthermore, we provide an extensive discussion on possible obstacles in traffic sign recognition tasks to facilitate future research through numerous experiments for each road condition.
Keywords:
traffic sign recognition; deep learning; object detection; real-time application; urban road scene
1. Introduction
Owing to the increasing market share of the autonomous vehicle industry, fundamental technologies for driving assistants and artificial intelligence have been increasingly studied [1]. One of the most crucial technologies for Advanced Driver Assistance Systems (ADAS), including self-driving, forward collision warning, or pedestrian recognition, is contextual awareness of road environments [2]. In particular, traffic sign recognition systems are core methods for providing vital instructions in safety-critical road regulations and should perform at highly stringent confidence levels.
Many autonomous vehicles utilize high-definition (HD) maps to provide richer information for road environments [3,4]. However, because of the manual and time-consuming efforts in production, the usage of HD maps is costly [5]. More importantly, HD maps can suffer from the discrepancies between the stored traffic signs and real-time changes [6]. In addition to assisting drivers, intelligent object recognition systems can facilitate the maintenance of road surroundings, such as traffic signs, lane lines, and guard rails [7]. For instance, traffic sign recognition systems can effectively analyze damage or defects through autonomous vehicles for monitoring purposes because it is nontrivial to inspect an entire road scene using human resources [8]. Therefore, the traffic sign recognition technique is an important component both for decision-making systems in vehicles
