Image-Based Real-Time Visual Tracking of Unknown 3-D Motion

This paper presents an image-based, real-time, visual servoing system capable of tracking unknown 3-D motion with high precision. The hardware system we used to achieve this purpose is The Real-time Intelligently ControLled Optical Positioning System (TRICLOPS). TRICLOPS mimics the foveal-peripheral capability of the human visual system by combining the visual sensing of a center camera with a wide angle lens (low resolution) with that of two vergence cameras with telephoto lenses (high resolution). The goal of tracking is to keep the object of interest in the optical axis of all three cameras. The rotation velocity input for the central camera system is obtained by the conventional image Jacobian method. The relative rotational positions of the two vergence cameras are obtained by taking the atan of the object position offset from the origin in the image plane over the focal length. Sub-pixel resolution can be obtained when the center camera and the two vergence cameras are combined in this fashion, especially when the object is close to the origin of the world frame.