Characterization of natural head and eye movements driving retinal flow
Characterization of natural head and eye movements driving retinal flow
复制标题
驱动视网膜血流的自然头部和眼睛运动的表征
DOI：
10.1167/19.10.147d
复制
发表时间：
2019
影响因子：
1.8
通讯作者：
Christian Sinnott
中科院分区：
文献类型：
--
作者：
P. MacNeilage;L. Nguyen;Christian Sinnott
在没有移动物体的情况下，视网膜血流由眼球相对于环境的速度以及环境的结构决定。空间中的眼睛速度是头在空间中的速度和在头中的眼睛速度的总和。为了更好地了解驱动视网膜血流的头部和眼睛速度，我们开发了一个系统，可以在实验室外的日常行为中同时测量头部和眼睛的速度。该系统由一个瞳孔实验室的眼睛跟踪器和一个固定在世界相机上的惯性测量单元(IMU)组成。头部速度是使用一种名为同时定位和映射(SLAM)的计算机视觉算法重建的，该算法通过跟踪场景中的特征来确定帧到帧的图像变形，然后求解产生该变形的相机运动。SLAM估计得到了来自IMU的角速度、线加速度和磁强计数据的补充。其结果是在20赫兹测量六自由度(6-DOF)头部速度，在120赫兹采样眼速。参与者在校园周围进行一系列活动时，记录了头部和眼球的速度。毫不奇怪，参与者倾向于注视静止环境的特征，强大的眼动稳定导致视网膜血流在中心凹附近微乎其微。视网膜血流的线性成分由头部的线速度驱动。然而，角分量并不强烈地依赖于角头部速度，因为角光流在很大程度上被补偿性眼球运动所抵消。取而代之的是，视网膜血流的角度成分是由注视时的线性光流补偿驱动的，这取决于注视相对于航向的偏心以及到场景的距离。因此，我们观察到视网膜血流受三个因素的影响最大：1)头线速度，2)注视方向和距离，以及3)环境的结构。
In the absence of moving objects, retinal flow is determined by eye velocity relative to the environment as well as by the structure of the environment. Eye velocity in space is the sum of head-in-space and eye-in-head velocity. To gain a better understanding of head and eye velocity driving retinal flow, we developed a system to measure both head and eye velocity during everyday behaviors outside the lab. The system consists of a Pupil Labs eye tracker with an inertial measurement unit (IMU) rigidly attached to the world camera. Head velocity is reconstructed using a computer vision algorithm known as simultaneous localization and mapping (SLAM) which works by tracking features in the scene to determine frame-to-frame image deformation, then solving for the camera motion that generated that deformation. The SLAM estimate is supplemented by angular velocity, linear acceleration, and magnetometer data from the IMU. The result is measurement of six-degree-of-freedom (6DOF) head velocity at 20 Hz with eye velocity sampled at 120 Hz. Head and eye velocity were recorded for participants performing a range of activities around campus. Not surprisingly, participants tend to fixate features of the stationary environment, and robust oculomotor stabilization leads to retinal flow that is minimal near the fovea. Linear components of retinal flow are driven by linear velocity of the head. Angular components, however, do not depend strongly on angular head velocity because angular optic flow is largely cancelled by compensatory eye movements. Instead, angular components of retinal flow are driven by compensation for linear optic flow at fixation, which depends on fixation eccentricity relative to the heading direction as well as distance to the scene. Consequently, we observe that retinal flow is driven most strongly by three factors: 1) linear head velocity, 2) fixation direction and distance, and 3) the structure of the environment.