Collaborative Research: Perception of 3D Objects in Dynamic Visual Scenes

人类观察者感知和识别三维场景中物体的能力涉及几个不同层次的分析，包括对场景整体深度的感知，对物体相对位置或场景布局的感知，以及对场景内物体形状的感知。在每个分析层次上，观察者可以从许多不同的来源获得信息，一些信息是关于观察者到部分场景或部分物体的距离（以观察者为中心），另一些信息是关于场景或物体内部的深度关系，与观察者的位置无关（以物体为中心）。将进行三个系列的实验，以确定场景、布局和对象等各个层面的信息，以及以观众为中心和以对象为中心的每种类型的信息如何结合起来，以提供对三维环境的综合感知。第一个系列的实验将检查模拟布局和感知布局之间的关系，因为场景深度的信息被操纵。第二组实验将评估布局信息对物体属性（相对深度、表面方向和整体形状）感知的影响。第三系列实验将使用前两个系列中选定的条件来检查场景深度和布局对物体属性感知的综合影响。在这些系列中，每个考虑中的关卡的信息类型（以观察者为中心或以对象为中心）将被操纵。以观众为中心和以对象为中心的信息规范的几个不同的线索将被检查，包括住宿，线性视角，运动视差，视差，相对大小，纹理，结构从运动和阴影。大多数实验将涉及计算机生成的显示器。还将进行一些涉及真实场景的实验。这将是对三维环境感知的多层次视觉信息整合的第一次全面研究。在观察者在场景中移动或必须识别或操纵场景中的物体的任何情况下，理解三维场景中物体的感知是很重要的。因此，在动态场景中人类视觉感知的基础研究与涉及人类和人工视觉系统的技术发展之间存在密切的关系，从改进飞行模拟器中飞行员训练的视觉显示到开发更有效的机器人视觉系统。＊＊＊

9510431 ANDERSEN The ability of human observers to perceive and recognize objects in a three-dimensional scene involves several different levels of analysis, including the perception of the overall depth of the scene, the perception of the relative positions of objects or layout of the scene, and the perception of the shapes of objects within the scene. At each level of analysis the observer may gain information from a number of different sources, some informative about distances from the observer to parts of the scene or to parts of objects (viewer-centered) and others informative about depth relationships within the scene or within objects, independent of the position of the observer (object- centered). Three series of experiments will be conducted to determine how information at each level, scene, layout, and object, and information of each type, viewer-centered and object- centered, combine to provide an integrated perception of the three-dimensional environment. The first series of experiments will examine the relationship between simulated layout and perceived layout as information for scene depth is manipulated. The second series of experiments will assess the effect of layout information on the perception of object properties (relative depth, surface orientation, and overall shape). The third series of experiments will use selected conditions from the first two series to examine the combined effects of scene depth and layout on the perception of object properties. Within each of these series, the type of information (viewer-centered or object-centered) for each level under consideration will be manipulated. Several different cues for the specification of viewer-centered and object-centered information will be examined including accommodation, linear perspective, motion parallax, disparity, relative size, texture, structure from motion, and shading. The majority of the experiments will involve computer-generated displays. Some experiments invo lving real scenes will also be conducted. This will be the first comprehensive study of the integration of visual information at multiple levels for the perception of the three-dimensional environment. An understanding of the perception of objects in three-dimensional scenes is important in any situation in which an observer moves through a scene or must recognize or manipulate objects in a scene. For this reason there is a close relationship between basic research on human visual perception in dynamic scenes and technological developments involving both human and artificial visual systems, ranging from improving visual displays for the training of pilots in flight simulators to developing more effective robotic visual systems. ***