Perception of Stereoscopically-Defined Surfaces and Motion

我们用眼睛来确定眼前场景的布局，并根据场景中静止和移动的障碍物来评估我们的运动路径的可取性。我们可以轻松地感知3d布局并进行视觉引导导航，这掩盖了这些任务的潜在复杂性。决定场景布局的主要刺激信息是双眼视差（例如，在魔眼立体图中，两只眼睛的图像产生深度感的差异）和运动视差（当一个人移动头部时，图像在视网膜上流动的速度的差异）。双目视差和运动视差在视网膜坐标中提供，但表面的位置和方向通常必须在以身体为中心的坐标中确定，以便指导四肢和躯干的适当动作。同样，判断物体运动方向（或自己运动方向）所需的信息包含在落在视网膜上的光的变化模式中，但运动方向通常必须相对于身体进行估计，以便于指导对场景中物体的运动反应。这项研究将研究两只眼睛图像的比较如何用于确定表面方向和曲率，以及确定物体运动的方向。这项研究将对我们理解人类空间感知很重要，但它也可能有实际的后果。首先，生物系统已经进化出强大的机制来估计表面的方向、形状和位置以及估计物体的速度。更好地理解生物系统的表现可能会导致更好的移动机器人系统算法。其次，感知物体相对于自身的运动对行动至关重要，包括对汽车的控制。因此，更好的理解可能会导致改进评估驾驶和飞行能力的程序。第三，更好地理解3D布局和运动感知中使用的视网膜和视网膜外信息可能有助于构建虚拟现实显示器。＊＊＊

9602080 BANKS We use our eyes to determine the layout of the scene before us and to assess the desirability of our path of motion with respect to stationary and moving obstacles in the scene. The ease with which we perceive 3-D layout and carry out visually-guided navigation belies the underlying complexity of these tasks. The primary stimulus information for determining scene layout is binocular disparity (the differences in the two eyes' images which creates the sensation of depth seen in, for example, the Magic Eye Stereograms) and motion parallax (differences in the speeds at which images stream across the retina when a person moves the head). Binocular disparity and motion parallax are provided in retinal coordinates, but the positions and orientations of surfaces must generally be determined in body- centered coordinates in order to guide appropriate actions of the limbs and torso. Similarly, the information needed to judge the direction of an object's motion (or one's own motion) is contained in the changing pattern of light falling on the retinas, but motion direction must generally be estimated relative to the body in order to be useful in guiding motor responses to objects in the scene. This research will examine how comparisons of the two eyes' images are used to determine surface orientation and curvature and to determine the direction of an object's motion. This research will be important to our understanding of human space perception, but it may well also have practical consequences. First, biological systems have evolved robust mechanisms for estimating surface orientation, shape, and position and for estimating object velocity. A better understanding of how biological systems perform may lead to better algorithms for mobile robotic systems. Second, the perception of an object's motion relative to the self is crucial for action, including the control of a motor vehicle. Thus, a better understanding could lead to improved procedures fo r assessing driving and flying capability. Third, a better understanding of the retinal and extra-retinal information used in the perception of 3D layout and motion may well aid construction of virtual reality displays. ***