Active camera tracking system using affine motion compensation

摘要

This paper describes a feature-based tracking system that can track moving objects with an active camera. A robust camera motion estimation algorithm is proposed to obtain a stable global motion for feature tracking. After we identify background and foreground motions based on dominant motion estimates, we estimate camera motion on the background by applying a parametric affine motion model. After compensating for camera motion, we trace multiple corner features in the scene and segment foreground objects by clustering motion trajectories of the corner features. We can also command the pan-tilt controller to position the moving object at the center of the camera.