Position Estimation Based Object Tracking across Multiple Cameras

An object tracking method based on position estimation across multiple cameras is proposed in this paper. For an object which is exiting a former view and entering in a latter, Kalman filter is used to predict the position in the next frame, and the reference region where the object will be observed in the latter view is obtained through a camera model. The object detected in the reference region is the correspondent object and the experiments show its validity.