An Improvement of the Camshift Human Tracking Algorithm Based on Deep Learning and the Kalman Filter

摘要

Automated human tracking in real time has been applied in many areas such as security, surveillance, traffic control, and robots. In this paper, an improvement of the Camshift human tracking algorithm based on deep learning and the Kalman filter is proposed. To detail an approach by using YOLOv4-tiny to detect a human in real time, Camshift is used to track a particular person and the Kalman filter is applied to enhance the performance of this algorithm in case of occlusion, noise, and different light conditions. The experiments show that the combination of YOLOv4-tiny and the improved Camshift algorithm raises the standard of speed as well as robustness. The proposed algorithm is suitable for running in real time and adapts well to the same color and different light conditions.