Perception of Layout and Motion during Self Motion in Real and Virtual Worlds

当我们在世界中移动时，我们的两只眼睛接收到一种复杂的视觉输入模式。局部图像速度随头部运动（计划和扰动）、场景布局和物体运动而变化。这种复杂的相对运动模式通常是根据光学阵列（视觉方向的流形）来建模的，光流描述了特征在光学阵列中的位置随时间的变化。虽然很复杂，但这种光流可以分解为由自身运动产生的流和由运动物体产生的流。当前提案的重点是更好地理解运动人类在动态环境中用于支持安全旅行的感知线索。然后，我将运用这些知识来帮助评估和开发更有效的虚拟现实3D用户界面。基于计算分析，我们将对人类用于支持运动的感知判断的视觉信息和参考框架进行预测。然后，我将设计实验性虚拟现实实验，以调查移动观察者面临的一些重要问题。这些包括(1)观察者如何将刚性环境的运动视差与障碍物、运动目标或干扰物的独立运动区分开来；(2)不严格约束于表面的特征（如镜面高光和阴影）的影响；(3)双眼周边视野在估计和控制航向、平衡和避障方面的作用是什么？这些问题的某些方面已经考虑了简单的刺激，但将扩展到考虑自然，双目虚拟和真实环境。拟议的研究将通过提供立体视觉在运动，定向和导航中的作用的关键见解来推进基础科学，这是一个尚未得到充分研究的主题。该提案将通过检查双目视觉在现实世界任务（如目标拦截或避障）中发挥的作用，扩展我们在场景感知和虚拟运动方面的工作。这些实验将由最先进的设备提供，如我们独特的宽视场立体显示器，移动眼动追踪和身体追踪，所有这些都与精心设计的行为和心理物理任务相结合。这项工作将以最先进的计算研究为基础，反过来，可以为计算机视觉提供生物学驱动的解决方案。高素质的研究人员将在加拿大工业界高度追捧的人的因素、人机交互和可视化技能方面接受培训。

When moving through the world our two eyes receive a complex pattern of visual input. The local image velocity varies with head motion (planned and perturbations), scene layout and object motion. This complex pattern of relative motion is usually modeled in terms of the optic array (manifold of visual directions) and the optic flow describes changes in the position of features in the optic array with time. While complex, this optic flow can be decomposed into flow arising from one's self-motion versus flow arising from moving objects. The focus of the current proposal is to better understand the perceptual cues that locomoting humans use to support safe travel in dynamic environments. I will then apply this knowledge to help evaluate and develop more effective 3D user interfaces for virtual reality. Based on computational analyses we will make predictions of the visual information and frames of reference used by humans to make perceptual judgements supporting locomotion. I will then design experimental virtual reality experiments to investigate a number of important questions faced by the moving observer. These include (1) how can an observer separate motion parallax of the rigid environment from the independent motion of an obstacle, locomotion target or distractor, (2) what is the influence of features that are not strictly bound to the surface such as specular highlights and shadows and (3) what is the role of the binocular peripheral visual field in the estimation and control of heading, balance and obstacle avoidance? Some aspects of these issues have been considered with simple stimuli but will be extended to consider naturalistic, binocular virtual and real environments. The proposed research will advance basic science by providing key insights into the role of stereoscopic vision in locomoting, orienting and navigating, a topic that has not been well studied. This proposal will extend our work on scene perception and virtual locomotion by examining the role that binocular vision plays during real world tasks such as target interception or obstacle avoidance. These experiments will be enabled by the availability of state-of the art facilities such as our unique wide-field stereoscopic display, mobile eye tracking and body tracking, all combined with carefully designed behavioural and psychophysical tasks. The work will be grounded in state-of-the-art computational research and could, in turn, inspire biologically motivated solutions for computer vision. Highly-qualified researchers will be trained on skills in human factors, human-computer interaction and visualization that are highly sought after by Canadian industry.