Tracking of Multiple Pixel Targets Using Multiple Cameras

Simultaneous tracking of multiple and small/pixel targets is important for many applications such as ship, aircraft, and vehicle tracking. This paper presents a practical and efficient framework for the above problem. The key ideas in the framework include 3D modeling of multiple cameras and the application of multiple Extended Kalman Filters (EKF) for multi-target tracking. Many practical issues such as multiple small targets, measurement noise, false targets, partial or missing target observations, multiple cameras, etc. have been addressed. Extensive experiments demonstrated the feasibility of the proposed framework.