灰度积分投影和圆形标记法结合的人眼定位

Gray Integral Projection Method Combined with the Circular Mark Eye Location Algorithm

讨论对于自然光下摄像头采集的人脸照片的眼睛定位算法,该算法是基于灰度积分投影和圆形标记法实现。分为以下三个步骤:首先,在RGB空间下对图像进行肤色检测,得到可能的人脸图像,通过形态学处理排除噪声等干扰;然后,根据亮度分量对人脸区域进行分割,得到五官图,并通过灰度投影对人眼进行粗定位;最后,通过圆形标记法,把区域内的白色空洞转化成面积相等且质心为圆心的圆形,经过几何特征筛选排除干扰圆得到双眼的两个圆形。该算法在Matlab平台上进行仿真实验,结果表明,此算法对于复杂背景下特别是存在类肤色干扰情况时人眼定位效果好、精度高。

Discussion the eye location algorithm of captureing the human face in photo by cameras on natural, which is based on the color information segmentation. Implementation is divided into the following three steps： First,for skin detection under the rbg space,the possible face images obtained,by morphological processing exclude noise interference; Then,based on the luminance component of the human face region segmentation,facial figure obtained; Finally,circular mark method,the area within the white empty into equal area and centroid as the center of the circle,through the geometric characteristics of filters eliminate interference circle got two round eyes. The algorithm in Matlab platform for simulation results show that the algorithm especially for complex background interference exists when skin color eye location is good,high accuracy.