ENGLISH
La vitrine de diffusion des publications et contributions des chercheurs de l'ÉTS
RECHERCHER

A refined traffic light detection method for challenging low visibility scenarios with CS-YOLOv10 algorithm

Zhang, Diyi, Zhou, Qinghui et Xie, Yidong. 2025. « A refined traffic light detection method for challenging low visibility scenarios with CS-YOLOv10 algorithm ». In Proceedings of the CSME-CFDSC-CSR 2025 International Congress (Montreal, QC, Canada, May 25-28, 2025) Coll. « Progress in Canadian Mechanical Engineering », vol. 8.

[thumbnail of 364 - A refined traffic light detection .pdf]
Prévisualisation
PDF
364 - A refined traffic light detection .pdf - Version publiée
Licence d'utilisation : Tous les droits réservés aux détenteurs du droit d'auteur.

Télécharger (484kB) | Prévisualisation

Résumé

Autonomous vehicles encounter significant challenges in accurately identifying traffic lights under adverse conditions such as fog and rain, due to reduced visibility and the difficulty of detecting small targets. To address these challenges, this paper presents an enhanced algorithm, KCS-YOLO, aimed at improving the detection and recognition of traffic lights in low visibility environments. Initially, we focused on improving the advanced YOLOv10 algorithm. To boost the detection of small targets, we developed the CS-YOLOv10 algorithm by incorporating the CA attention mechanism and creating a dedicated small target detection layer. Additionally, we created a traffic light image dataset and applied the dark channel prior dehazing algorithm for preprocessing to enhance the recognition accuracy and robustness of the proposed algorithm. Finally, the CS-YOLOv10 algorithm was evaluated through comparison and ablation experiments. The results demonstrated that the mAP of the CS-YOLOv10 algorithm reached 98.32%, a 5.68% improvement over the YOLOv10 baseline. These findings confirm that the CS-YOLOv10 algorithm offers high accuracy in object detection and recognition, significantly enhancing traffic light detection for autonomous vehicles in low visibility conditions.

Type de document: Compte rendu de conférence
Éditeurs:
Éditeurs
ORCID
Hof, Lucas A.
NON SPÉCIFIÉ
Di Labbio, Giuseppe
NON SPÉCIFIÉ
Tahan, Antoine
NON SPÉCIFIÉ
Sanjosé, Marlène
NON SPÉCIFIÉ
Lalonde, Sébastien
NON SPÉCIFIÉ
Demarquette, Nicole R.
NON SPÉCIFIÉ
Date de dépôt: 18 déc. 2025 15:34
Dernière modification: 18 déc. 2025 15:34
URI: https://espace2.etsmtl.ca/id/eprint/32530

Actions (Authentification requise)

Dernière vérification avant le dépôt Dernière vérification avant le dépôt