A Stereo Vision-based Guidance System for the Visually Impaired People

This paper proposes a stereo vision-based algorithm for objects’ depth estimation suggested for guiding visually impaired people. The proposed algorithm is a first step towards a complete system designed for warning visually impaired persons to avoid in front obstacles at a reasonable distance (1m-2m). First, the left and right images are captured by a stereo vision camera. After preprocessing the captured images, objects are detected and segmented using morphological operations then a stereo image rectification is performed. The inliers points differences obtained from epipolar constraint are used to calculate the maximum and the minimum disparity. The suggested algorithm in this paper is tested using images captured for different objects with various distances from the stereo camera. Experimental results show that the proposed algorithm for depth estimation works quite robustly and faster than other methods in the literature.