Characterizing Dimmable Augmented Reality for Highlighted Target Detection

Author(s):

Roman A. Stemp and Chou P. Hung

Abstract:

Augmented reality can enhance visual search performance, but the luminance contrast of target highlighting can be overly distracting, impairing search performance in real-world environments. Real-world scenes, particularly with combinations of shadow and strong illumination, can vary in luminance up to 100,000:1 in a single view, making it difficult to control the highlighting contrast to avoid attentional capture, impairing the ability to detect unhighlighted targets. We characterized how target detection performance is affected by a novel target highlighting technique called divisive display augmented reality (ddAR), in which a region is highlighted by dimming the transparency versus adding luminance. Across a choice of six possible target locations, we tested varying levels of target and highlight contrast and different numbers of targets. We also tested conditions where the highlighting is missing or misleading, simulating different types of aided target recognition malfunction. We report that 20% dimming significantly increased target detection, with low cost to the detection rate of a second unhighlighted target. Lower highlight contrasts (2%, 5%, and 10% dimming) had little-to-negligible effect on detection compared with baseline (0%), whereas higher contrasts (40%, 60%, and 100% dimming) became too visually salient, producing an attentional capture effect distracting from unhighlighted targets.