Hybridization of FA-YOLOv8 Model with Zebra Optimization Algorithm for Improving Traffic Light Detection and Tracking Under Low-Light Conditions

Traffic-light detection and tracking under low-light conditions remain difficult because signal heads can be small, dim, blurred, low in contrast, or affected by glare and reflections. This study develops a hybridization of Firefly Algorithm–You Only Look Once version 8 (FA-YOLOv8) with the Zebra Optimization Algorithm (ZOA) for offline analysis of recorded dashcam video under rainy, night-time, and cloudy conditions. YOLOv8 detects traffic-light candidates, FA performs attraction-based coordinate refinement and supports tracking continuity, and ZOA acts as an enhancement layer that adaptively searches FA control settings rather than replacing FA as the refinement mechanism. A controlled within-dataset comparison uses 300 real-world dashcam frames containing approximately 900 annotated traffic-light instances; the same annotated sequences support parameter adaptation and the baseline-versus-hybrid comparison, so the reported results characterize within-dataset behavior rather than held-out generalization. Compared with baseline FA-YOLOv8, MOTP increases from 31.84% to 37.59% under rainy conditions and from 49.47% to 72.50% under cloudy conditions. At night, Recall increases from 54.70% to 80.34% and F1-Score from 61.73% to 75.56%. Tracking continuity also improves in the reported point estimates: IDF1 rises from 72/90/85% to 82/94/90% for rainy/night/cloudy conditions, while identity switches fall from 95/18/45 to 48/8/22. These observations show that ZOA-based enhancement changes FA search behavior and is associated with improvements in several localization and identity-tracking measures within the evaluated sequences, although detection changes remain condition-dependent.

@article{Salemi2026,
  title     = {Hybridization of FA-YOLOv8 Model with Zebra Optimization Algorithm for Improving Traffic Light Detection and Tracking Under Low-Light Conditions},
  journal   = {International Journal of Advanced Computer Science and Applications},
  volume    = {17},
  number    = {9},
  year      = {2026},
  publisher = {The Science and Information Organization},
  author    = {Adhwa Salemi and Muhammad Arif Mohamad},
  doi       = {10.14569/IJACSA.2026.0170955},
  url       = {https://doi.org/10.14569/IJACSA.2026.0170955}
}