HCC: Egocentric Depth Perception in Augmented Reality

增强现实（AR）是一种计算机显示器将计算机生成的图形对象添加（叠加）到用户对物理（真实）世界的视图中的技术。AR与更为人所知的虚拟现实（VR）不同，在VR中，观察者看到的是完全由计算机生成的图形场景。AR使现实世界中没有对应的可视化技术成为可能；其中一项技术是x射线视觉，增强现实用户可以感知位于固体、不透明表面后面的物体。在这个项目中，PI将实证研究以自我为中心的深度感知（观察者到物体的距离）如何在AR中运作。PI将进行一系列实验，在这些实验中，观察者判断AR呈现的虚拟物体的深度。这些实验将使用两种不同类型的依赖度量：视觉定向操作和基于应用程序的任务。一种常用的视觉指导行动是盲行，观察者看到一个目标，遮住眼睛，然后在没有视觉的情况下走到目标位置。有很好的理论论据表明，视觉导向的行为测量的是相对纯粹的自我中心距离感知，不受观察者认知知识的污染。此外，有大量的经验数据描述了现实世界物体的视觉定向动作距离判断，以及用VR显示设备观看的虚拟物体。这些数据表明，在全线索条件下，对现实世界物体的判断没有系统误差，误差可达~20米，而VR物体的距离被系统地低估了（这一现象已经被广泛研究，但尚未完全解释）。基于应用程序的任务是由AR技术的引人注目的应用程序驱动的。其中一项任务是感知匹配，即虚拟物体和真实物体的深度匹配；这项任务是AR在医学和图像辅助手术、城市环境和建筑物中的AR态势感知等方面应用的重要组成部分。另一个这样的任务是强制选择，其中虚拟对象的深度被放置到相对于其他对象的少数类别之一。这项任务是由AR机场控制塔和AR城市态势感知系统等应用驱动的，在这些应用中，观察者必须根据虚拟和真实物体的总体空间布局做出决策。尽管对现实世界物体和vr呈现的虚拟物体的自我中心深度感知已经被广泛研究，但目前关于这一问题的经验数据很少，本研究将纠正这一缺失。此外，由于目前的研究将使用两种互补的依赖度量，它们将允许测量诸如VR低估效应等现象的程度，这种现象已被视觉指导的行动任务发现，也存在于定性不同的依赖度量中。这将有助于解决有争议的问题，即这种现象在多大程度上源于依赖测量与更深层次的感知机制的选择。更广泛的影响：在应用环境中，为了实现许多引人注目的AR应用，更好地理解AR深度感知是如何运作的是必要的，通过这项活动收集的经验数据将加速AR应用的开发。此外，通过这项活动，一系列学生将获得计算机图形学和人类实验方法的综合经验；毕业后，这些学生将很好地为计算机图形学应用感知这一重要的新兴研究领域做出贡献。

Augmented reality (AR) is a technology where computer displays add (superimpose) computer-generated, graphical objects to a user's view of the physical (real) world. AR is distinguished from the better-known virtual reality (VR), wherein an observer sees an entirely computer-generated graphical scene. AR makes possible visualization techniques that have no real-world equivalent; one such technique is x-ray vision, where AR users perceive objects which are located behind solid, opaque surfaces. In this project the PI will empirically study how egocentric depth perception (the distance from an observer to an object) operates in AR. The PI will conduct a series of experiments, in which observers judge the depth of AR-presented virtual objects. These experiments will use two different categories of dependent measures: visually directed actions, and application-based tasks.A commonly used visually directed action is blind walking, where observers view a target, cover their eyes, and then walk to the target location without sight. There are good theoretical arguments that visually directed actions measure a relatively pure percept of egocentric distance, uncontaminated by observers' cognitive knowledge. Furthermore, there is a substantial body of empirical data that describes visually directed action distance judgments of both real-world objects, and virtual objects viewed with VR display devices. These data indicate that, under full-cue conditions, real-world objects are judged without systematic error up to ~20 meters, while the distance of VR objects is systematically underestimated (a phenomenon which has been studied extensively but not yet fully explained).The application-based tasks are motivated by compelling applications of AR technology. One such task is perceptual matching, where the depth of a virtual and a real object are matched; this task is an important component of AR applications in medicine and image-assisted surgery, AR situation awareness in urban settings and buildings, and others. Another such task is forced choice, where the depth of a virtual object is placed into one of a small number of categories relative to other objects. This task is motivated by applications such as an AR airport control tower and an AR urban situation awareness system, where observers must make decisions based on the gross spatial arrangement of virtual and real objects.Although the egocentric depth perception of real-world objects and VR-presented virtual objects has been widely studied, currently there exists very little empirical data on the issue, an absence this research will correct. Furthermore, because the present studies will use two complimentary categories of dependent measures, they will allow measuring the degree to which phenomena such as the VR underestimation effect, which has been found by visually directed action tasks, is also present in qualitatively different dependent measures. This will help resolve controversial questions regarding the degree to which such phenomena arise from the choice of dependent measure versus deeper perceptual mechanisms.Broader Impacts: In an applied context, a better understanding of how AR depth perception operates is necessary for many compelling AR applications to be realized, and the empirical data gathered through this activity will hasten AR application development. In addition, through this activity a series of students will receive a blend of experience in both computer graphics and human-subject empirical methods; upon graduation these students will be well-positioned to contribute to the important emerging research area of applied perception in computer graphics.