We present the architecture of a system for tracking multiple targets, using multiple cameras with possibly overlapping fields of views, to determine the 3D location and orientation of the targets. A separate tracker is assigned to each new target. That tracker uses the information provided by 2D pattem matchers about the location of the target in the cameras'' video streams to update its estimate of the target''s 3D state. An extended Kalman Filter (EKF) is used to maintain each target''s 3D state. New targets are detected by employing an adaptive background subtraction algorithm. We present examples of results of our tracker for outdoor scenes.
展开▼
