Improving the Accuracy of Depth Estimation in Binocular Vision for Robotic Applications

Abstract

Binocular vision or stereo vision for extraction of three-dimensional information from stereo images has been widely used in many applications like robot navigation, recovering the three-dimensional structure of a scene, and optical inspection systems. More recently, the majority of research in binocular vision has focused on the establishment of stereo matching. However, to date, there has been relatively little research conducted on the effect of computational models of binocular vision with variable focal length of lens. In this paper, a modified computational model of binocular vision is presented to develop a new depth estimation algorithm with no effect of changes in focal length. This method provides an obvious advantage in accuracy of depth estimation by reducing the effect of changing the lens focal length. The experimental results show that the proposed depth estimation method in binocular vision provides better accuracy than conventional method. Finally, we apply the new depth estimation method to a stereo-vision-based automatic docking system for a mobile robot to verify its accuracy.