Object Recognition and Navigation

About 3.7 million Americans are visually disabled. Of these, 200,000 are blind, and the rest have low vision. There has been almost no research on low-vision problems in object recognition or navigation. Our main goal is to use the principles of visual science to understand and ameliorate these problems. We believe that object recognition and navigation are conceptually similar with respect to the effects of vision loss. 

Our Proposal has two theoretical goals. The first is to extend theory on normal vision to encompass low vision. We will do this using ideal-observer analysis in which low vision is characterized as having sensory constraints beyond those of normal vision. Our success in applying ideal-observer analysis to reading and low vision encourages us to undertake a similar theoretical approach in object recognition and navigation. Second, active exploration is often involved in object recognition in the form of eye movements or magnifier movements, and in navigation in the form of head and body movements. This is especially true in low vision where field loss from disease or field restrictions from magnifiers limit the portion of an image or scene that can be viewed at one time. By developing "active ideal-observer" models, we hope to address an important theoretical problem: how do we account for the strategies used by people with normal and low vision in exploring objects or spatial layouts? 

The main empirical goal of Series 1 in the Research Plan is to measure and characterize eye-movement exploration and magnifier exploration in object recognition. When image features are hard to see, due to poor visibility, impaired vision, or magnifier limitations, do people employ effective strategies for finding and using informative features? What kind of visual features do they use? In Series 2, we ask an important practical question: Can cognitive maps serve as useful navigation aids for people with low vision? We will conduct experiments in virtual and real buildings. Our long-term goal is to develop a computer-based method for imparting information about building layouts to visually disabled people. In this grant, we will study the impact of impaired vision, especially field loss, on the formation and use of cognitive maps. Our hypothesis is that impaired vision will make it more difficult to acquire cognitive-map knowledge, but once acquired, this knowledge will be especially beneficial to people with low vision. 

Our research team brings three areas of particular strength to this proposal: computational modeling using ideal-observer analysis, expertise on low vision, and expertise on visuomotor performance and spatial cognition.