Hybrid firefly algorithm and you only look once v8 framework (FA-YOLOv8) for traffic light detection
Accurate traffic light detection is a critical component of intelligent transportation systems and autonomous driving. However, existing deep learning approaches struggle under adverse conditions such as occlusion, illumination variation, and cluttered environments. This study introduces a novel hyb...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Conference or Workshop Item |
| Language: | en |
| Published: |
IEEE
2026
|
| Subjects: | |
| Online Access: | https://umpir.ump.edu.my/id/eprint/46465/1/Hybrid%20firefly%20algorithm%20and%20you%20only%20look%20once%20v8%20framework.pdf https://umpir.ump.edu.my/id/eprint/46465/ https://doi.org/10.1109/ICSECS65227.2025.11279255 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Accurate traffic light detection is a critical component of intelligent transportation systems and autonomous driving. However, existing deep learning approaches struggle under adverse conditions such as occlusion, illumination variation, and cluttered environments. This study introduces a novel hybrid framework that integrates the Firefly Algorithm (FA) with You Only Look Once version 8 (YOLOv8), termed FA-YOLOv8, to enhance accuracy and precision in traffic light detection. The proposed FA-YOLOv8 was tested on three video datasets, comprising a total of 340 frames and 1,684 objects, which in each dataset captured in different challenges (occlusion, illumination and cluttering). As a result, the proposed hybrid FA-YOLOv8 method achieved a high precision of 96.76%. In comparison with three existing approaches which are YOLOv5+DeepSort, YOLOv8+DeepSort, and ByteTrack, the proposed hybrid FA-YOLOv8 method demonstrated superior performance, achieving the highest scores in five out of ten evaluation metrics, including Multiple Object Tracking Precision (MOTP), Recall, F1-Score, (Identity F1) IDF1 and (Mostly Lost) ML. Its ease of use and ability to function in both offline and real-time scenarios make it versatile and suitable for various domains. Applications include traffic surveillance, vehicle monitoring, and intelligent transportation systems, such as autonomous driving. Detecting traffic light accurately and efficiently is essential for effective vehicle regulation and roadway safety. This research contributes a valuable framework for improving detection methods across diverse real-world environments and advanced transportation technologies. |
|---|