Dynamic object tracing method for mobile robot by fusing visual features

This paper proposes a novel method of tracing dynamic objects for indoor mobile robot by fusing visual features. The proposed tracking algorithm is a kind of improved CAMSHIFT algorithm combined with SURF, optical flow and ultra-sonic data, to enhance the tracking capability. Since CAMSHIFT algorithm is based HSV color space model, it is weak on strong lighting, abrupt change of light and fast-moving object. Also, CAMSHIFT doesn't consider the object's moving direction and velocity information when it sets the search window for tracking. To set a searching window for tracking, the visual features of SURF, optical flow and ultra-sonic data are used. So, we can analyse the size, moving direction, velocity of the fast-moving objects. Experimental tests prove the improved efficiency performance compared with the conventional CAMSHIFT only.