IMAGO: Image-guided navigation for visually impaired people

Abstract

Blind and visually impaired persons face many challenges due to isolation, the most important is lacking education due to social immobility. Yet, since the development of the well-known white-red cane, only few advances have been made to increase the mobility of blind people. GPS systems are often used for pedestrian navigation but lack precision, foremost in rural areas where navigation is most often needed. The aim of the IMAGO project is therefore the development of an inexpensive and unobtrusive navigation method for blind and visually impaired persons. Navigation is performed using structure from motion and image-based localization techniques. Route models are created as 3D point clouds through several steps: (i) image acquisition along the routes; (ii) bulk-transfer of images to a server; (iii) feature extraction; (iv) feature matching between images; (v) creation of 3D point cloud with structure from motion. Similarly, the navigation chains seven steps: (a) image acquisition while walking a route; (b) immediate transfer of image to a server; (c) feature extraction; (d) feature matching between image and 3D route model; (e) localization/camera matrix calculation; (f) navigation/calculation of direction based on localization; (g) transfer of direction to user. Current smartphones are used as devices both for recording of routes as well as navigation. Thereby, a high level of dissemination without additional costs is possible, both, within blind people for navigation, as well as seeing people for route creation. Additionally, haptic feedback can be used via a smart cane to reduce auditive feedback. The proposed system yields a high positioning accuracy of 80% of samples being located within 1.6 m. Thus, the system is usable for pedestrian navigation, especially for visually impaired persons.