Research on Traffic Sign Area Detection and Classification Method Based on Improved YOLOv8s
Journal of Engineering System · 2026-09 · 4 (3) · 113-118
DOI: 10.62517/jes.202602311 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Engineering System, 2026-09. Verfasst von Feng Liu und 3 weiteren Autor:innen. Die Arbeit nennt 10 Literaturangaben.
Abstract (Originalsprache)
A detection and classification method based on improved YOLOv8s is proposed to address the problems of low recognition rate of traffic signs in complex environments, easy missed detection of small targets, and insufficient real-time performance. By integrating GTSRB, TT100K public datasets, and autonomously collected images, a dataset containing 45 categories and approximately 12000 samples was constructed through cleaning, enhancement, and unified annotation. Using YOLOv8s as the baseline, introduce CIoU loss function and adaptive anchor box clustering to optimize the confidence loss weight. The experiment shows that the improved method mAP@0.5 It reaches 73.9%, which is 7.1 and 8.0 percentage points higher than YOLOv8n and Faster R-CNN, respectively, with an inference speed of 120ms/frame. The confusion matrix and F1 curve validate the classification accuracy and anti-interference ability of the model, which can provide effective technical support for autonomous driving and intelligent traffic management.
Zitieren
APA
Liu, F., Guo, W., Zhao, Y., & Li, Y. (2026). Research on Traffic Sign Area Detection and Classification Method Based on Improved YOLOv8s. Journal of Engineering System, 4(3), 113-118. https://doi.org/10.62517/jes.202602311
Vancouver
Liu F, Guo W, Zhao Y, Li Y. Research on Traffic Sign Area Detection and Classification Method Based on Improved YOLOv8s. Journal of Engineering System. 2026;4(3):113-118. doi:10.62517/jes.202602311
BibTeX
@article{Liu2026,
author = {Liu, Feng and Guo, Weiwei and Zhao, Yunlong and Li, Ying},
title = {Research on Traffic Sign Area Detection and Classification Method Based on Improved YOLOv8s},
journal = {Journal of Engineering System},
year = {2026},
volume = {4},
number = {3},
pages = {113--118},
doi = {10.62517/jes.202602311}
}
EUVIMED