Assistive guidance system for the visually impaired
- Published
- Accepted
- Subject Areas
- Human-Computer Interaction, Artificial Intelligence, Autonomous Systems, Computer Vision
- Keywords
- Visually impaired, Obstacle detection, Crosswalk detection, Staircase detection, Raspberry Pi, OpenCV
- Copyright
- © 2017 Takhar et al.
- Licence
- This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, reproduction and adaptation in any medium and for any purpose provided that it is properly attributed. For attribution, the original author(s), title, publication source (PeerJ Preprints) and either DOI or URL of the article must be cited.
- Cite this article
- 2017. Assistive guidance system for the visually impaired. PeerJ Preprints 5:e3410v1 https://doi.org/10.7287/peerj.preprints.3410v1
Abstract
In recent years, with the improvement in imaging technology, the quality of small cameras have significantly improved. Coupled with the introduction of credit-card sized single-board computers such as Raspberry Pi, it is now possible to integrate a small camera with a wearable computer. This paper aims to develop a low cost product, using a webcam and Raspberry Pi, for visually-impaired people, which can assist them in detecting and recognising pedestrian crosswalks and staircases. There are two steps involved in detection and recognition of the obstacles i.e pedestrian crosswalks and staircases. In detection algorithm, we extract Haar features from the video frames and push these features to our Haar classifier. In recognition algorithm, we first convert the RGB image to HSV and apply histogram equalization to make the pixel intensity uniform. This is followed by image segmentation and contour detection. These detected contours are passed through a pre-processor which extracts the region of interests (ROI). We applied different statistical methods on these ROI to differentiate between staircases and pedestrian crosswalks. The detection and recognition results on our datasets demonstrate the effectiveness of our system.
Author Comment
This is a preprint submission to PeerJ Preprints.