International Journal of Advanced Research in Computer and Communication Engineering

Abstract: The rapid growth of urbanization and the increasing number of two-wheelers on roads have significantly intensified traffic congestion and safety challenges. Traditional traffic monitoring systems rely heavily on manual observation or basic image processing techniques, which often fail to provide accurate, real-time analysis under complex road conditions. To overcome these limitations, this project presents a computer vision–based traffic object detection system using YOLOv8, designed specifically for bike-only traffic scenarios. The proposed system focuses on detecting critical traffic-related objects, including person, helmet, and vehicle number plate, from surveillance video streams. Video frames are extracted using OpenCV and manually annotated using LabelMe. Since YOLOv8 does not support JSON annotations directly, the annotations are converted into YOLO format with normalized bounding box coordinates. A pretrained YOLOv8 model is fine-tuned on the custom dataset to achieve accurate real-time detection. During inference, the trained model processes video frames and outputs bounding boxes with class labels and confidence scores. Experimental results demonstrate reliable detection performance under both daytime and nighttime conditions, with minimal false detections. The modular architecture of the system enables easy extension for higher-level traffic analysis such as helmet violation detection and number plate recognition.The proposed approach provides an efficient, scalable, and intelligent solution for automated traffic surveillance and serves as a strong foundation for smart transportation systems

Keywords: Traffic Object Detection, YOLOv8, Computer Vision, Helmet Detection, Number Plate Detection, Intelligent Transportation System

Downloads: | DOI: 10.17148/IJARCCE.2026.15161