Target tracking with an attentive foveal sensor

Abstract

This paper addresses the problem of tracking a moving target using measurements taken by a foveal sensor; i.e., a sensor having a foveal region of high acuity surrounded by a periphery in which the acuity of measurements is much less. As in human vision, the foveal region is assumed to be movable and the algorithms developed attempt to exploit it by keeping the target within the foveal region as much as possible. When tracking a target whose motion is modeled by the state of a linear system driven by white gaussian noise, an approach for management of the foveal sensor motion is demonstrated that allows the foveal sensor to always outperform a sensor having no fovea (i.e., all periphery) and to approach the performance of a sensor having no periphery (i.e., all fovea) in some cases.