An Efficient Method of Tracking across Multiple Cameras

Tracking people across multiple cameras is a challenging task in video surveillance system. Field of view (FOV) lines were introduced as an efficient tool to resolve the consistent labeling issue. We introduce a new and robust method to find the overlapping FOV of two cameras, and the concept of homography is applied to find the correspondence of the moving object in the two views. This approach does not rely on the object movement in the scene as proposed in earlier techniques. The labeling scheme suggested in the past was based on the distance of an object from an FOV line is improved. The solid FOV concept makes the process less prone to errors and computationally efficient. This method can be easily applied to any multi-camera system to efficiently track the moving object in different cameras and FOV of one camera in another camera view can be drawn.