This project aims to explore, develop and test computer vision algorithms to analyze images of street intersections from a camera worn by a blind person. Urban intersections are the most dangerous parts of a blind person's travel. They are becoming increasingly complex, making safe crossing using conventional blind orientation and mobility techniques ever more difficult. We will explore computer vision algorithms to help a blind person find the crosswalk, find the pedestrian signal button, determine when the """"""""walk"""""""" light is on, and alert him/her to any veering out of the crosswalk. We will emphasize the development of completely novel methods of analyzing non-ideal images including shadows, occlusions and other irregularities using spatial grouping techniques based on Bayesian inference. The resulting algorithms are intended for eventual integration as modules for a computer vision system we are already developing to help blind persons with travel tasks such as finding and reading aloud printed signs and negotiating street crossings. The combined system would have potential for a radical advance in independent travel for blind persons. In this exploratory project, we aim to: (1) Explore and test alternative approaches to algorithm design to process intersection images and extract the information about the crosswalk, crossing signal, etc., using a database of real-world images taken by blind persons at a variety of different kinds of intersections. (2) Test the algorithms using a portable camera connected to a notebook computer with speech output.

Agency
National Institute of Health (NIH)
Institute
National Eye Institute (NEI)
Type
Exploratory/Developmental Grants (R21)
Project #
1R21EY015187-01A2
Application #
6920594
Study Section
Special Emphasis Panel (ZRG1-BDCN-F (12))
Program Officer
Oberdorfer, Michael
Project Start
2005-08-01
Project End
2007-07-31
Budget Start
2005-08-01
Budget End
2006-07-31
Support Year
1
Fiscal Year
2005
Total Cost
$255,198
Indirect Cost
Name
Smith-Kettlewell Eye Research Institute
Department
Type
DUNS #
073121105
City
San Francisco
State
CA
Country
United States
Zip Code
94115
Shen, Huiying; Coughlan, James; Ivanchenko, Volodymyr (2009) Figure-Ground Segmentation Using Factor Graphs. Image Vis Comput 27:854-863
Shen, Huiying; Chan, Kee-Yip; Coughlan, James et al. (2008) A mobile phone system to find crosswalks for visually impaired pedestrians. Technol Disabil 20:217-224
Ivanchenko, Volodymyr; Coughlan, James; Shen, Huiying (2008) Crosswatch: a Camera Phone System for Orienting Visually Impaired Pedestrians at Traffic Intersections. Lect Notes Comput Sci 5105:1122-1128
Ivanchenko, Volodymyr; Coughlan, James; Shen, Huiying (2008) Detecting and Locating Crosswalks using a Camera Phone. Proc IEEE Comput Soc Conf Comput Vis Pattern Recognit 2008:4563143
Renninger, Laura Walker; Verghese, Preeti; Coughlan, James (2007) Where to look next? Eye movements reduce local uncertainty. J Vis 7:6