On the Design and Structure of Artificial Eyes for Tracking Tasks

Abstract

Nature developed different types of eyes for different functions and creatures. In this work the design parameters to common eye types are analysed with respect to visual tracking abilities. The goal is to select optimal parameters for building artificial vision systems. The parameters investigated are different tessellations and resolutions of eye cells (pixels), the window size or field of view, the frequency to acquire visual input and the system latency or reaction time. The analysis shows that following a target is best achieved at high frequency with small target windows or with a space-variant tessellation. While the first case enables high resolution for smooth motions, the second case trades resolution for speed and jerky motions. An experimental set-up demonstrates the performance characteristics.