Wearable Computing for Image-Based Indoor Navigation of the Visually Impaired

Includes bibliographical references (pages 27-29)In this thesis, an image-based non-obtrusive indoor navigation system for the visually impaired is presented. The system makes use of image processing algorithms to extract floor regions from images captured from a wearable eye-mounted heads-up display. A prototype system called VirtualEyes is presented, where floor regions are analyzed to provide the user with voiced guidance for navigation. The floor detection algorithm was tested against over 200 images captured from indoor corridors of various lighting conditions and achieved up to 81.8% accuracy. The system proves to be an effective approach to navigational guidance for the visually impaired