Pedestrian Tracking Using YOLOv8 with Lucas-Kanade Optical Flow for Traffic Light Control Application

A robust pedestrian tracking system using YOLOv8 for detection and Lucas-Kanade optical flow for motion tracking under varying lighting conditions is developed. Pedestrian tracking is essential for surveillance, robotics, and autonomous vehicles, requiring accurate detection in dynamic environments. A Raspberry Pi camera is employed for real-time processing, and the system's performance is evaluated in artificial and real-world lighting scenarios. Results show a chi-square statistic value - 0.081, significantly lesser than the critical value of 5.991. Results suggests - no significant association exists between the lighting conditions. The results demonstrate effective pedestrian tracking capabilities, highlighting the system's potential to enhance pedestrian safety and traffic management. A significant challenge to be addressed is mitigating human occlusion, which is a primary problem for tracking accuracy in crowded or obstructed environments. Future research should focus on advanced techniques to minimize occlusion effects, ensuring reliable performance across diverse real-world conditions and validating their practical applicability.