Multiple human tracking using PHD filter in distributed camera network

Abstract

The Gaussian mixture probability hypothesis density (GM-PHD) filter is a closed form approximation of the multi-target Bayes filter which can overcome most multitarget tracking problems. Limited field of view, decreasing cost of cameras, and advances of using multi-camera induce us to use large-scale camera networks. In this paper, a multihuman tracking framework using the PHD filter in a distributed camera network is proposed. Each camera tracks objects locally with PHD filter and a track-after-detect scheme and its estimates of targets are sent to neighboring nodes. Then each camera fuses its local estimates with it's neighbors. The proposed method is evaluated on the public PETS2009 dataset. The results measured in Correct Tracking Percentage (CTP) showed a better performance compared to one of the most recent related works on the evaluated dataset.