Blind Cam Problem Smartphone wayfinding for the blind
Blind Cam Problem Smart-phone way-finding for the blind Kee-Yip Chan, kyc@soe. ucsc. edu Roberto Manduchi, manduchi@soe. ucsc. edu James Coughlan, coughlan@ski. org UC Santa Cruz SURF-IT, Summer 2007 User launches program to start locating color target. 2. Pitch/volume reports aiming accuracy. 3. If aim is good inform user, “Target Located, ” and begin searching/decoding barcode. Blind people have difficulty wayfinding in office environments. • They must find assistance or stumble by themselves. Alternative solutions Proposed solution • User interaction 1. • • • Cheap resources, real-time response, and inexpensive management. • Symbian C++ different from standard. • No exception handling. • Asynchronous modules, schedulers. Blind Cam project modules: • Camera (Async) • Audio (Async) • • Color. Target. Engine Wireless systems are expensive to manage. • GPS does not work indoors. Using an identifiable label (barcode) and smart-phone with built-in camera. Color target localization in real-time. • Camera rotation tolerance, distance tolerance. Computer Vision techniques to find label and convey information back to user. • Camera exposure affects shutter speed, limited configurability. • White balance affects color consistency, limited configurability. Aiming accuracy Color target • Based on color gradients, advantage over textured patterns. • Empirically determined color combination, arrangement. • Robust to orientation of camera, lighting conditions. • Algorithm: cascade filter, comparing color gradients from color patches. • • • Finding adjacent color target (fiducial) instead of barcode improves localization speed. TTS (Async) User must first locate Braille signs. Limitations and performance Symbian S 60 programming • • Minimal computation cost. • Horizontal error from captured image center. • Inform user of aim through pitch/volume changes. Distance R 1, G 1, B 1 R 3, G 3, B 3 R 2, G 2, B 2 • Inferred from size of color target in captured image. • Algorithm: Ellipse fitting, seeded region growing.
- Slides: 1