EYE MOVEMENTS IN THE PERCEPTION OF DEPTH FROM MOTION

从运动中感知深度可能是视觉处理的重要方面。单眼的深度线索都很容易被欺骗，而双眼立体视觉对于单眼观察者或眼睛侧向放置的动物来说是不可用的。然而，来自运动的深度可能产生模糊的深度分配（例如，动态深度效应- kde）或在运动视差的情况下产生明确的深度感知，即使视网膜信息在这两种情况下可能是相同的。是什么让运动视差产生的深度清晰？虽然关于观察者头部运动方向的信息似乎是必要的，但目前的研究表明，视觉系统实际上使用缓慢的眼动信号来实现这一目的。这个项目将研究视觉系统完成这项任务的机制。心理物理实验将使用能够整合线性头部追踪系统和红外眼球追踪系统信息的计算机系统，这样就可以监测头部和眼睛的位置。该系统还可以呈现运动视差显示的立体版本（使用铁电快门系统），以便双目立体视觉可以用作比较运动视差深度的控制。实验将确定OKR或追求眼球运动在从运动中感知深度的效用，包括先前模糊的条件，如KDE。此外，由于眼动幅度取决于观看距离，因此将探讨观看距离在运动视差显示中的作用。最后，如果运动视差的深度依赖于缓慢的眼球运动，那么由于皮质或小脑病变导致的眼动功能障碍患者应该在运动视差的深度感知方面存在问题。理论目标是开发一个单一的神经网络模型，该模型可以从运动视差、运动视角、运动深度、立体视觉以及这些不同深度线索之间的各种神经相互作用来解释深度感知。以前的神经网络模型（Nawrot和Blake， 1991）局限于动力学深度、立体视觉及其相互作用。眼动信号的加入将扩大模型的范围，包括运动视差及其与双目立体视觉的各种相互作用。

The perception of depth from motion is perhaps the important aspect of visual processing. Monocular cues to depth are all easily deceived, and binocular stereopsis is not available to monocular observers or animals with laterally places eyes. However, depth from motion may produce either ambiguous depth assignments (e.g., kinetic depth effect-KDE) or unambiguous depth percepts in the case of motion parallax even though the retinal information may be the same in the two cases. What is it that makes depth from motion parallax unambiguous? While information about the direction of observer head movement appears necessary, current research shows that the visual system actually uses a slow eye movement signal for this purpose. This project will study the mechanism by which the visual system accomplishes this task. Psychophysical experiments will use a computer system capable of integrating information from a linear head-tracking system and an infra-red eye-tracking system so that both head and eye position can be monitored. This system can also present stereoscopic versions of the motion parallax displays (using ferroelectric shutter system) so that binocular stereopsis can be used as a control to which depth from motion parallax is compared. Experiments will determine the utility of OKR or pursuit eye movements in the perception of depth from motion, including previously ambiguous conditions such as KDE. Additionally, as eye movement magnitude depends on viewing distance, the role of viewing distance in motion parallax displays will be explored. Finally, if depth from motion parallax relies on slow eye movements, then patients with eye movement dysfunction due to cortical or cerebellar lesions should show problems in the perception of depth from motion parallax. The theoretical goal is to develop a single neural network model that can account for the perception of depth from motion parallax, motion perspective, kinetic depth, stereopsis and the various neural interactions between these different depth cues. The previous neural network model (Nawrot and Blake, 1991) was confined to kinetic depth, stereopsis, and their interaction. The addition of an eye movement signal would expand the scope of the model to include motion parallax and its various interactions with binocular stereopsis.