中文摘要：

针对大型人脸数据库中进行人脸匹配识别时存在识别速度时间长、影响实时应用效果的问题,提出了一种基于凸包的人脸粗分类新方法。该方法从几何模式特征出发,以抽取人脸的二维凸包不变量特征为基础,使用层次聚类对人脸的轮廓线进行粗分类,建立人脸数据库的层次索引结构。在实验中,将MUCT和PICS人脸数据库的正面人脸图像粗分为六类,分类的平均准确率约为89%。验证了该方法在人脸数据库上执行快速粗分类是可行的。

英文摘要：

The face recognition system involves large amounts of data matching operation under the condition of massive faces database. It will give rise to problem such as long retrieval time and can＇t match the real-time requirement. To solve the problems above,this paper proposed a new method for face coarse classification based on convex hull,which started from the geometric features of pattern,used the two-dimensional convex hull invariant features to extract the face contour lines. It then classified the contour lines to several subclasses through hierarchical clustering method. Finally,it divided the massive faces database into a number of sub databases resulting in hierarchy architecture,and classified the frontal face images in MUCT and PICS face databases into six classes in the experiment by the proposed method respectively. The average accuracy rate of classification is about 89%. The experiment also verifies the feasibility of the proposed method for face coarse classification in massive faces database.