Authors: [tex2html_wrap4224]T. Baron, M. D. Levine, Y. Yeshurun (Tel Aviv University, Israel)
Investigator username: levine
Subcategory: sensor and processor design
We have been developing a biologically-motivated robot eye system FOVIA for active vision. Our implemented system computes a foveated image (fovea + periphery) from which it determines appropriate gaze control inputs to move the sensor and focus its attention on the salient parts of the scene. To demonstrate this concept, we have formulated various types of attentional operators which are used to select salient interest points which drive the gaze control mechanism. Our system efficiently integrates images from space-variant snapshots of an arbitrary natural environment. Foveational control permits us to build and maintain a global and integrated long term representation of the scene from a set of image acquisitions. In future, we plan to use the system to study new foveated gaze control strategies such as task-oriented focus of attention, movement detection, tracking and motion analysis.