Augmented reality performance assessment battery (ARPAB): Object recognition, distance estimation and size estimation using optical see-through head-worn displays.

摘要

Augmented reality (AR) is a maturing technology that often utilizes head-worn see-through displays and shows digital objects so they appear as part of the environment. AR systems are increasingly being used in training and to support on-the-job performance. As a result, interface designers will increasingly be faced with developing user interfaces for these systems with minimal knowledge of the human factors of their use.; This study examined human factors issues underlying the use of AR systems by investigating people's ability to identify objects, judge distances, and estimate sizes of objects. Specifically, the study used real objects, 3 dimensional (3D) models of those real objects, basic 3D geometric shapes, and flat geometric shapes (e.g., chair, VRML model of a chair, a pyramid, and a triangle).; The participants wore off-the-shelf head-worn displays, either a Sony Glasstron (LCD display) or a Microvision Nomad (virtual retinal display). They viewed objects displayed individually at distances from 10–110 feet and reported estimated distances, size estimates, and object identity.; The basic design of the experiment was a Test (4) by Distance (4) by Repetitions (2), where each of these factors was a fully crossed within-subject factor. The factors were randomized and counterbalanced using a Latin square design. In addition, there were several nested factors within tests. The primary data analyses consisted of Repeated-Measures ANOVAs dealing with within-subjects factors. After inspection of the control data, little variance was seen. Thus, to preserve error degrees of freedom, these covariates were not included in the analyses.; Results confirm the importance of environmental cues for depth perception. People performed better and were more confident in their estimates with real objects than with virtual objects. Estimates for 3D models of real objects were more accurate than 3D geometric shapes. In turn, they estimated 3D geometric shapes more accurately than flat geometric shapes. Also, on all measures, participant error decreased with distance. These results indicate a need to greatly increase the image display quality of head-worn displays before commercial augmented reality systems are deployed. Additionally, people are likely to need extensive training in how to best judge depth of the objects displayed.