Continuous Location Tracking of People by Multiple Depth Cameras

This paper describes the design and implementation of a novel multi-depth camera multi-person tracking system. We propose a zone-space model for effectively cooperating with multiple depth cameras. Based on the system model, we devise a tracking-session association algorithm to find out a correct mapping between new observations and users scattered in the zones consisting of a space. Our system model supports flexible and extensible dynamic management of zones consisting of a space. Experiment results show that the proposed tracking-session algorithm shows good tracking performance such as location errors and hand over between zones.