UC Santa Cruz

Visually impaired people are a group of people who have some degree of vision loss. In order to help these people getting text information of surrounding space, I build an app that can detect text for visually impaired people. At first, the app takes images automatically, and then the google OCR API is used to find text spots and perform word recognition. After that, my app will notify the user of the text contents and positions. Since there is overlapping between adjacent images, the app might detect one text twice if the text exists in both photos. To solve this problem, homography matrices among pictures are calculated using feature detection and matching techniques. Besides, the panorama is used to illustrate my results. As a result, the app can detect text efficiently and reduce various difficulties that visually impaired people might meet in walking, navigating and so on.