Real time object tracking based on retino-cortical mapping

Abstract

Human vision has two areas: one is fovea centralis which is located at the center of visual axis and has high resolution, and the other is the peripheral where the resolution decreases according to the distance from the fovea. This relationship between distance and visual resolution is known as the log polar mapping, which realizes two different functions in central and peripheral visions, respectively. The object is focused and recognized with high resolution in the central vision, but the area is small. However, the objects in a large area are roughly recognized in peripheral vision, with low resolution. We propose the new method of target tracking with log polar mapping adapted to sequential image and with moving-area extraction through the optical flow obtained by the difference image method and generalized gradient-based method. Through our system, the target tracking is realized with high resolution, during short period of processing time, with data processing reduced.