Towards Saliency-based Gaze Control in a Binocular Robot Head
André Filipe Roos, Hugo Vieira Neto · 2015
Abstract – Robots that emulate biological visual systems must solve the perceptual problem of when and where to direct gaze. In this paper, a gaze control scheme for a specific binocular robot head is described and assessed. A well-known computational model of visual attention is used to find attractive target locations and a proportional-integral position control is used to drive pan and tilt servomotors of a dominant camera. Experiments demonstrate that the system is capable of tracking conspicuous moving targets in real-time (30 fps) in non-cluttered environments without any previous knowledge or strong assumptions. Addition of vergence control for the non-dominant camera is the next step towards a complete binocular control framework.