Image-based visual tracking to fast moving target for active binocular robot

Abstract

The paper reports the control way of visual tracking to a fast moving target with using an active binocular robot. The robot consists of a pair of active cameras and performs as a stereo vision system. The motions of those cameras are controlled independently and achieve rigid visual tracking respectively. The difficulty of visual tracking to a fast target is to suppress the tracking delay, affected by the high-speed of the target and the rapid changes of its motions. Although we have established an explicit theory for non-delayed visual tracking in our previous works, it is fact that the good tuning of the control parameters should realize its own high performance. In this paper, we have examined the effective tunings of three types of the control parameters; the feedback gains for motion control and for visual servoing and the image capturing frequency. In several conditions, the higher performance on the visual tracking to a fast moving target has been achieved. The validity of the proposed method is confirmed in the many physical experimental results.