Intelligent gaze control in binocular vision

Robotic vision systems that incorporate multiple cooperating sensors, mounted on maneuverable platforms, and their control are addressed. Assuming binocular, maneuverable (approximately anthropomorphic) computer vision hardware, the authors explore how sensorimotor control algorithms can contribute to visual tasks and behaviors. The authors discuss issues in the organization of robotic gaze stabilization, the implementation and application of vergence in binocular systems, and the use of nonvisual cues in stabilizing gaze. Stabilizing its gaze can enable an animate vision system to interpret and interact with its environment more effectively. Binocular cues and vergence contribute a precategorical segmentation of an object of interest that permits gaze stabilization to be preattentive and therefore very general. Nonvisual cues offer the potential for improving gaze stabilization performance by enabling the system to sense and compensate for its own motion, thereby allowing the visual signals to describe primarily target motion