Identification and tracking of overlapping vehicles by high-speed vision

Abstract

In this paper, we propose high-speed tracking system using machine learning and image processing to identify overlapping vehicles. Since existing tracking system can track only one car and fail to track overlapping cars, it is necessary to be able to identify overlapping cars and continue tracking. We adopt deep learning algorithm YOLO[1] to identify overlapping cars and high-speed tracking system using correlation filter and IPM method for tracking. Our proposed method implemented high-speed and high-precision tracking system for overlapping cars.