Improving Computer Vision-Based Indoor Wayfinding for Blind Persons with Context Information

There are more than 161 million visually impaired people in the world today, of which 37 million are blind. Camera-based computer vision systems have the potential to assist blind persons to independently access unfamiliar buildings. Signs with text play a very important role in identification of bathrooms, exits, office doors, and elevators. In this paper, we present an effective and robust method of text extraction and recognition to improve computer vision-based indoor wayfinding. First, we extract regions containing text information from indoor signage with multiple colors and complex background and then identify text characters in the extracted regions by using the features of size, aspect ratio and nested edge boundaries. Based on the consistence of distances between two neighboring characters in a text string, the identified text characters have been normalized before they are recognized by using off-theshelf optical character recognition (OCR) software products and output as speech for blind users.