利用测地距离的三维人脸定位算法

Algorithm of 3D Face Location Using Geodesic Distance

针对传统的二维人脸定位,无法克服旋转、表情、姿态等因素带来的问题,同时传统定位算法的准确率较低,算法在三维人脸模型的基础上,加入测地距离,提出利用测地距离的三维人脸定位算法。首先输入待检测的三维人脸图像,对其进行维纳滤波预处理,在预处理后的图像中进行鼻尖点定位,进而找到人脸的位置,在待检测图像中标记所得到的人脸区域。算法在三维人脸库FRGC和BU-3DFE上进行实验,利用深度信息定位方法和SPIDER特征点定位方法进行对比,实验结果表明本算法的定位准确率更高,鲁棒性更强。

The traditional two-dimensionnal face positoning unable overcome the rotation,expression,posture,and own a low accuracy in the location.We join the geodesic distance on the3D face modle and proposed the algorithm of three-dimension location using geodesic distance.We use wiener filtering to preprocess the3-dimentional face datas for the detecting image and confirm the location of the face by finding the location of nose point in the preprocessing image called nose tip location.Then we unify the human faces to the same coordinate frame.Finally,and mark the face region to be detected in the resulting image.The algorithm makes the experiment on FRGC face database and BU-3DFE face database,uses the depth information positioning method and spiders feature point positioning methods for comparing.The experimental results show that our algorithm of positioning accuracy is higher,stronger and has good robustness.