Motorcycle-related traffic violations, including helmetless riding and multiple riders, present significant challenges for urban traffic management, especially under conditions of occlusion and high traffic density. This research introduces a hybrid, real-time traffic violation detection framework that improves detection reliability by combining deep learning-based object detection with a secondary verification mechanism. The framework utilizes the lightweight YOLOv8n model to detect motorcycles, riders, and helmets, and incorporates a Haar Cascade-based module to verify rider count and minimize false detections in congested environments. Evaluation is conducted using real-world traffic video data collected across diverse environmental conditions. Performance is measured using standard metrics such as precision, recall, F1-score, and mean average precision (mAP@0.5). Experimental results show that the proposed method achieves a detection accuracy of 93.4% and a mAP of 94.8%, while sustaining real-time processing at approximately 38 frames per second. The verification stage enhances detection robustness, particularly in cases of occlusion and overlapping riders. These findings suggest that the hybrid framework offers a reliable and computationally efficient solution for real-time traffic violation detection, with strong applicability in intelligent transportation systems and smart city initiatives.
Bashir et al. (Mon,) studied this question.