Tracking a Person with Pre-recorded Image Database and a Pan, Tilt, and Zoom Camera

This paper proposes a novel tracking strategy that can robustly track a person or other object within a fixed environment using a pan, tilt, and zoom camera with the help of a pre-recorded image database. We define a set called the Minimum Camera Parameter Settings (MCPS) which contains just enough camera states as required to survey the environment for the target. This set of states is used to facilitate tracking and segmentation. The idea is to store a background imageof the environment for every camera state in MCPS, thus creating an image database. Duringtracking camera movements are restricted to states in MCPS (or a version of this set that is augmented to improve smoothness of tracking). Scanningfor the target and segmentation of the target fromthebackgroundaresimplifiedaseachcurrent image can be compared with the corresponding pre-recorded background image.

• Type: Conference
• Year: 1998
• Where: ACCV
• Authors: Yiming Ye, John K. Tsotsos, Karen Bennet, Eric Harley
• Keywords: ACCV 1998, Camera, Camera State, Computer Vision, Image Database