A Review of Real-Time Intelligent Traffic Light Control Systems: Bridging Gaps with Multi-Camera GPU-Accelerated YOLO for Heterogeneous Traffic

Abstract

This paper presents a critical review of existing intelligent traffic light control systems, advanced computer vision techniques, and You Only Look Once (YOLO) object detection applications. It highlights the identified research gap concerning the integrated application of multi-camera parallel processing, GPU-accelerated frame processing, and fine-tuned YOLO models for real-time three-wheeler detection in the context of intelligent traffic light control. The paper concludes by discussing the significance of addressing this gap for enhancing urban mobility and safety in heterogeneous traffic environments. A notable study introduced a multi-camera YOLOv8-based real-time system incorporating GPU acceleration and a dedicated threewheeler dataset, achieving competitive mAP scores with efficient identity tracking across cameras. This method also addressed viewpoint variance, occlusion, and heterogeneous traffic conditions by deploying parallel frame processing pipelines using Jetson Orin devices for edge inference [1] Recent advancements in bridge traffic monitoring have emphasized the value of vision-based systems. Previous work introduced an open-source dataset specialized for vision-based traffic load monitoring on cable-stay bridges. This benchmark dataset includes over 32,000 annotated images, calibration data, and synchronized video-WIM recordings, enabling deep learning model training and validation in real bridge environments. The dataset supports accurate spatial load distribution estimation and encourages further development of bridge traffic analysis technologies [2].

Description

Citation

Prameeth, S. (2025) A Review of Real-Time Intelligent Traffic Light Control Systems: Bridging Gaps with Multi-Camera GPU-Accelerated YOLO for Heterogeneous Traffic, Technology Innovations & Digitalization Accompanying Creativity (TIDAC), NSBM Green University, Sri Lanka. P.338

Endorsement

Review

Supplemented By

Referenced By