Binocular vergence control using disparity energy neurons

Abstract

The human vision system accounts for a major part of the cortex and has been a subject of interest for investigators of 3-D perception. Vergence control is a crucial component for robust and consistent binocular vision. In turn, such visual components provide for attention-based saccadic observations. Robust vergence control can be realised using the disparity energy model inspired by the study on disparity-tuned neurons in the mammalian visual cortex and it has been considered the most biologically plausible model for disparity estimation. While the method provides a faithful model for disparity estimation, the computational complexity of the model impedes the use of it in real-time robotic systems. This article proposes a hierarchical disparity estimation model using fixed tuning disparity energy cells and a pyramidal image structure to restructure the concepts of the coarse-to-fine disparity energy model so that near real time vergence control can be achieved. Such is the achievement of this article, that scientific perspectives can be used to fulfil engineering level objectives through the use of biologically inspired cognitive studies. This improvement of speed has led to a successful deployment of a binocular vision system capable of vergence in a natural indoor environment. Also presented are the theories and results of the vergence control and the disparity estimation, including a systematic presentation of both the qualitative and quantitative results.