Baseline Independent Binocular Vergence Control of 2 DOF Pan-Tilt Cameras using a Visual cortical Model

Abstract

CogniVision is a neurologically inspired vergence control model that uses the optimization of the disparity error between interlaced, log-polar transformed, cortical maps incident on the visual cortex (VC) [13]. This paper presents results that illustrate a biologically inspired vergence control of a pair of 2 DOF pan-tilt cameras. This work illustrates the ability of the cortical model to perform binocular vergence with baseline independence and absence of object registration. It was inspired by a collection of discoveries in the area of neuroscience. See Schwartz (1977) [1], Tootell et al (1982) [2] and Nobel Prize Winner Hubel et al (1974) [4]. We briefly explain how the log- polar image incident on the VC can be used successfully to determine the pan angle (saccade) to perform binocular vergence on points of in the scene. A discussion at the end of the paper highlights the significant differences between this system and the conventional object correspondence systems that require matching of specific object shapes. In this model, the cortically magnified VC image is used to match the entire image.