Rui Jin¹ (), Gordon Legge², Roberto Manduchi³, Yingzi Xiong¹; ¹Wilmer Eye Institute, Johns Hopkins University, ²Department of Psychology, University of Minnesota, ³Department of Computer Science and Engineering, University of California, Santa Cruz

Low vision individuals usually require closer viewing distance than those with normal vision for object recognition. We define critical viewing distance (CVD) as the minimum distance at which an individual can correctly recognize an object, a crucial determinant for safe and effective navigation. In real-world navigation, objects are often viewed dynamically with gradually decreasing distances due to self or object motion, rather than statically at fixed viewing distances. Whether such dynamic viewing aids object recognition compared to static viewing, especially with degraded vision, is not well understood. Ten participants were asked to recognize common navigation icons (e.g., no crossing, no parking) on a computer display. The icons were chosen through a pilot study to ensure similar difficulty and filtered digitally to simulate low vision with 0.8 logMAR and 0.3 logCS. To measure CVD, the icons expanded in size, mimicking a gradual decrease in viewing distance from 7.2m to 1m by a constant speed (0.7m/s). Participants were asked to stop the trial as soon as they could recognize an icon, and CVD was computed as the equivalent viewing distances. Participants also performed two static viewing conditions: brief viewing (500ms) and unlimited viewing, at five equivalent viewing distances from 5m to 1m, with recognition accuracy obtained at each distance to derive psychometric curves. Across participants, CVD under dynamic viewing averaged 2.93m with a recognition accuracy of 96.5%, which was significantly higher than both static conditions (65.4% and 90.6%, ps < .01) at each participant’s CVD level. Modeling results showed that such benefits of dynamic viewing could not be explained by probability accumulation of brief viewing samples (p < .001). Our findings provide initial evidence on the importance of dynamic viewing for optimizing object recognition. Ongoing studies are expanding to complex objects, under self-motion, and with actual low vision in real world navigation.