The manhunt network: People tracking in hybrid-overlapping under the vertical top-view depth camera networks
An-Sheng Liu, Tang-Wei Hsu, Po-Hao Hsiao, Yen‐Cheng Liu, Li‐Chen Fu · 2016
The importance of surveillance system increases in recent years. The ability of camera tracking has also gone mature enough to produce reliable results. But even so, the automotive tracking algorithm in multiple views is still a late bloomer. In this paper, we are exploring a general form of cameras' coalition tracking. We not only propose a new 3D biometric feature for identification in camera networks, but also focus on reformulating the framework of a camera network. Here, a new hybrid-overlapping coalition of a set of cameras is proposed for either overlapping or non-overlapping tracking. In the meantime, a novel time-varying non-overlapping strategy inspirited by the manhunt procedure of law enforcement is developed. The proposed approach is combining the overlapping and nonoverlapping (hybrid-overlapping) camera networks together to track targets collaboratively for the better efficient.