基于三维采样点集的人脸识别

Face Recognition Based on Sampled 3D Points Clouds

采用基于整体轮廓的提取方法来对三维人脸点集进行重采样表征人脸。首先将三维人脸区域的点集校正到统一的姿态坐标系,并将其转换为深度图,之后计算深度图的一阶和二阶梯度,并设定阈值提取出边界曲线,再找出二维梯度图的边界曲线所对应的三维空间中的曲线点集用来表征人脸,最后用D-ICP算法进行配准并进行相似度测量。在欧洲人脸数据库GAVAB3D中进行了测试,实验结果表明该方法简便有效。

This paper resampled the 3Dface points clouds based on its whole contour to characterize the face.Firstly,the points set of 3Dface region will be adjusted into a unified attitude coordinates and then be changed into depth images.Secondly,the first and second order of gradient ratio of the depth images will be calculated,also a certain threshold will be set to extract the boundary curves of them,in succession,the corresponding points clouds in 3Dspace of the 2D boundary curves will be found to characterize the human face.Finally,D-ICP algorithm will be used in model registration step and also the similarity of two models will be measured.The experiments were carried out on the European face database GAVAB3 D.And the results indicate that our method is handy and effective.