Enhanced Speed Estimation Based on 2D Object Detection and Monocular Vehicle Pose Estimation

This study aims to enhance the accuracy of vehicle speed estimation in camera-based Intelligent Transportation Systems (ITS). By integrating advanced technologies, including YOLOv7 object detection, ByteTrack multi-object tracking, and Ego-Net monocular vehicle pose estimation, we have improved the reliability of speed estimation results. Experimental findings demonstrate that our novel framework significantly increases the accuracy of estimation, providing trustworthy speed estimation results. The study validates the feasibility of this framework, offering experimental results highly consistent with ground truth GPS data and providing robust support for ITS research and applications.