A minimal architecture for detecting object location and motion

摘要

This paper comprises a novel architecture for constructing a simple vision system for motion detection and tracking. The idea is based on a mechanism of the fly visual system. Its elements employ the simplest possible oscillatory activation function and are parameter free. Furthermore, its elements are hierarchically organized, which means that motion can be detected at a local or a global level. Simulation results are described which show that the system is able to track a moving stimulus, and extrapolate the current motion of a stimulus. Finally, it is shown that even in a noisy environment the system is able to detect a moving stimulus.