Visual Servo Tracking of a Wheeled Mobile Robot Utilizing Both Feature and Depth Information

This paper presents a novel visual servo tracking method for a wheeled mobile robot based on target features and depth of scenery. The robot is equipped with an RGBD sensor to provide both color and depth information of the scene. The proposed method comprises two phases. The first phase utilizes feature matching to ensure that the target moves into the field of vision such that the robot may enter the second phase. The second phase implements visual servo aiming at object tacking accuracy. The depth information is employed in the proposed approach, which reduces the computational complexity required as in conventional approaches based on feature extraction. In addition, the second phase has built-in mechanism for detecting and exiting local minimum when the robot speed converges. The method is also capable of sequentially tracking multiple objects. Experimental results demonstrate the feasibility and effectiveness of the proposed two-phase method.