Approximately one in ten blind persons are confined to wheelchairs, and independent travel is currently next to impossible for this population. Conventional blind wayfinding techniques - cane or guide dog - become extremely difficult or impractical in a wheelchair, requiring great physical dexterity and coordination. The traveler's loss of direct contact with the ground also removes vital feedback. It is easy to miss hazards such as obstacles and drop-offs ahead of or alongside the chair. While some technologies have been developed to aid in wheelchair navigation, they are either restricted to controlled environments or detect only a limited range of obstacles very close to the chair. As a result, independent travel is so difficult that few attempt it, resulting in a widespread lack of awareness of this severely disadvantaged population. The PI's long-term goal is to make independent travel feasible for blind wheelchair users. In this project, he will tackle the problem of sensing terrain features that are relevant to wayfinding in a wheelchair using computer vision technology. Computer vision algorithms for interpreting visual scenes will be developed to infer important visual information, in real time, about nearby terrain, obtained from images collected by cameras mounted to the wheelchair. This information will include the detection of hazards such as obstacles and drop-offs ahead of or alongside the chair, as well as detecting veer, finding curb cuts, finding a clear path, and maintaining a straight course. It will be communicated to the traveler using synthesized speech, audible tones and/or tactile feedback, and is meant to augment rather than replace the information from existing wayfinding skills. The traveler will use this information to control the wheelchair himself/herself (rather than relying on robotic control of the chair). The outcome will be a prototype system that helps prevent veering off the sidewalk by performing the following functions: detect and locate drop-offs and other obstacles; detect and locate curbs and curb cuts; and detect and locate the shoreline (i.e., edge of the sidewalk bordering grass or other terrain, or adjoining a wall), as well as sideslopes.
Broader Impact: Beyond the tremendous benefits that the technology resulting from this research may impart to the population of blind wheelchair users, the research will also advance the understanding of computer vision algorithms applied to terrain analysis and interpretation. The unique research environment at Smith-Kettlewell will draw upon the expertise of in-house blind engineers (one of whom is a post-doctoral fellow), who will assist in the design and assessment of the proposed system. In addition, the study will provide an opportunity for minorities and blind persons to gain exposure to science while participating in the gathering of data as research subjects. Finally, the results of the research will be disseminated in several public forums, including demonstrations of the system at local high schools and at the Exploratorium Science Museum of San Francisco.
