On performance enhancement of a following tracker using stereo vision
Nam-Gyu Oh, Jae Il Cho, Kiheon Park · 2010
A stereo vision system has a distinctive advantage capable of obtaining exact and wide range of the distance information. The use of stereo vision system has been increased due to the advances in hardware technology. This paper deals with design of the active vision system which acquires the distance and the direction information of the target using the stereo vision system. This paper also implements the tracker that follows the main tracker, which we call a following tracker, without identifying the target. In addition, the CAMShift algorithm which tracks the target based on the color information is used to acquire the center point of the object. A following tracker algorithm generates the control input that makes the following tracker aim the same target based on the sensing results from the active vision system. A following tracker needs the exact distance information between the system and the target to track the object accurately. We first build the mathematical model of object tracking system and following tracker, design the system controllers, and then compare the tracking accuracy under the two conditions: the presence and the absence of the distant information.