基于FCM聚类和C-V模型的人眼精确定位方法

Precise human eyes location algorithm based on FCM Clustering and C-V model

提出了一种精确提取眼睛轮廓、虹膜边界、瞳孔中心及眼角坐标的方法。首先利用基于AdaBoost算法的眼睛检测器提取眼睛所在的图像子区域,并应用单尺度Retinex算法增强该子区域;然后在子区域中运用梯度Hough圆变换提取瞳孔中心和虹膜边界;接着运用模糊C均值(FCM)聚类算法在子区域内提取眼睛的初始区域,并以初始区域构造符号距离函数作为C-V模型的初始水平集函数;最后运用C-V模型提取眼睛轮廓和眼角坐标。在Purdue AR人脸测试集上,结合FCM聚类后C-V模型的收敛速度提高了64.1%、定位精度提高了8.3%。方法不受复杂背景影响,对光照变化有较好的适应度,具有较高的鲁棒性。实验结果表明方法是有效的。

In order to precisely locate the human eye features,such as eye contours,iris boundaries and the coordinates of pupil centers and canthi,a hierarchical approach is presented.First,an eye detector is trained by AdaBoost algorithm for extracting the sub-images containing the eyes,and the sub-images are enhanced by single-scale Retinex algorithm.Second,the pupil centers and iris boundaries are located using gradient Hough circle transform in the sub-images.Third,the eye regions are segmented by FCM clustering for constructing the initial signed distance function which is the Level Set function of C-V Model.Finally,the eye contours and canthi are located using C-V Model.With the combination of FCM clustering and C-V Model,the convergence rate of C-V Model is increased by 64.1% on the Purdue AR face test set,while the locating accuracy is increased by 8.3%.This algorithm is immune to complex background and illumination changes,showing that it has high robustness.Results show that the proposed approach is efficient.