一种眼动测量的图像分析方法

提出一种用于鸟类视觉行为学实验的眼动测量的图像分析方法.采用基于遗传算法的多级灰度值聚类法分割普通CCD捕捉的视频图像,用区域生长法标记连通域粗略定位瞳孔区域,并利用瞳孔的近似圆形的几何特性修补光源反射影像形成的孔洞.在此基础上进行边缘检测,利用边缘像素的灰度分布特点修正瞳孔轮廓,采用重心法定位瞳孔中心.用该方法对实验环境照度下捕捉的图像和红外光源辅助照明的图像进行了分析,并与主动轮廓线方法对比.实验结果表明,该方法对眼睛特征的先验知识依赖程度低,抗噪声能力强,瞳孔中心定位精确.

An Image Analysis Method for Ocular Movement Measurement

For the avian vision study in behavioral experiments, an image analysis method is developed in order to measure the ocular movement of Aves based upon captured images from common CCD cameras. Firstly, genetic algorithm (GA) based multi-level gray clustering is used to segment the original image, and region growing is used to label the consecutive areas so as to rudely locate the pupil region. Secondly, the holes in pupil caused by cornea reflection are eliminated according to the approximate circinal geometric property of the pupil. Finally, the edge of the pupil is detected using the mean and standard deviation of the histogram of the edge pixels to correct the profile of pupil, and then the center of gravity is calculated to determine the pupil center. The proposed method is applied to the visible-light image and IR-illuminated image in the experimental condition, and contrasted with a method based on active contour. The results show excellent performance in the respect of less dependence on the prior knowledge of eye features, the robustness for noise and the precise location of the pupil.