Collision-Free Navigation for Blind Persons using Stereo Matching

Saeid Fazli, Hajar Mohammadi D. Abstract— A blind person in an urban environment has to navigate around obstacles and hazards. Though a significant amount of work has been done on detecting obstacles, not much attention has been given to the detection of drop offs, e.g., sidewalk curbs, downward stairs, and other hazards where an error could lead to disastrous consequences. In this paper, we propose algorithms for detecting obstacles in an urban setting using stereo vision and Two-Stage Dynamic Programming (TSDP) technique. We are developing computer vision algorithms for sensing important terrain features as an aid to blind navigation, which interpret visual information obtained from images collected by cameras mounted on camera legs nearly as high as young person. This paper focuses specifically on a novel computer vision algorithm for detecting obstacles, which are important and ubiquitous features on and near sidewalks and other walkways. A fast and robust stereo matching algorithm is proposed that uses only features in order to find a semi-dense disparity map. It works by growing from a small set of correspondence seeds. In this paper we use Normalized Cross-correlation (NCC) function matching with a 5 × 5 window and prepare an edge detected matching table τ and start growing disparity components by drawing a seed s from S which is computed using Sobel edge detector, and adding it to τ. It results in high accuracy and performance. The obstacle identified in the proposed method which appears in the disparity map enters the phase of depth computing. The algorithm extracts as much information as possible from depth information obtained from a stereo system. The proposed method is tested on a Pentium 4 Dual CPU 2.00 GHz PC with 2.00 GB of RAM and achieves better RMS error and bad matching percentage compared to previous methods. Experimental results are demonstrated on typical sidewalk scenes to show the effectiveness of the proposed method.