RGB-D Information Based Multi-people Tracking and Monitoring for a Sports Assisting System

The rapid development of deep learning and computer vision technology has brought about revolutionary changes in the field of people tracking. In particular, the tracking of athletes’ movements in sports scenes such as basketball and soccer has been attracting attention as an important research theme. The problem with conventional tracking algorithms is that they usually detect false positives or mistakes in dynamic environments such as fast player movements, dense players, or complex interactions. In this research, we propose a multi-target tracking and monitoring system with high accuracy in a dynamic environment by combing the deep-learning based tracking model with human identification algorisms based on RGB-D information. The effectiveness of the proposed system is shown by actual experiments.