Automated Seat Belt Detection Using Surveillance Cameras for Indian Traffic

Computer Engineering

Road accidents continue to rise at an alarming rate across India, highlighting the urgent demand for robust traffic safety enforcement systems. Among various preventive measures, seat belts remain one of the most effective means of reducing crash-related injuries, yet widespread non-compliance persists as a significant concern. This paper introduces an automated deep learning-based framework designed to identify seat belt violations within Indian traffic scenarios. The system leverages lightweight YOLOv8 model variants to perform vehicle detection, directional classification, and windshield-level analysis to assess whether occupants are wearing seat belts. To address performance degra- dation under adverse lighting and environmental conditions, Contrast Limited Adaptive Histogram Equalization (CLAHE) is integrated into the pipeline, yielding a 5.7% improvement in detection accuracy. The proposed system attains a precision of 87.5% and a recall of 86.5% for seat belt detection, with an inference speed of 13.9 ms per frame, demonstrating its viability for real-time traffic monitoring and enforcement applications. This work advances the field of intelligent transportation systems by offering a scalable and efficient solution toward safer urban road environments.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.