融合改进的ASM和AAM的人脸形状特征点定位算法  

Facial feature point localization algorithm combining improved ASM and AAM

作　　者：王晶[1] 苏光大[1] 刘炯鑫[1] 任小龙[1] 

机构地区：[1]清华大学电子工程系,北京100084

出　　处：《光电子．激光》2011年第8期1227-1230,共4页Journal of Optoelectronics·Laser

基　　金：公安部重点攻关计划资助项目(2005ZDGGQHDX005);国家"973"计划资助项目(2007CB310600)

摘　　要：基于主动形状模型(ASM)和主动表面模型(AAM),提出了一种融合改进的ASM和AAM的人脸形状特征点定位算法。利用ASM定位外轮廓的形状特征点,AAM定位内部形状特征点;采用对部分关键特征点使用二维梯度的方法以提高特征点搜索的准确性;利用眼、鼻和嘴这些关键特征点的定位信息初始化人脸的平均形状以改善初始位置不当造成的搜索失败;建立多尺度的ASM以提高收敛速度。实验结果表明,本文方法比传统的ASM、AAM方法以及已有的改进算法IASM和PAAM定位更精确。Active shape models （ASM） and active appearance model （AAM） are the primary methods of facial feature point localization. In this paper,an improved facial feature localization algorithm that combines ASM and AAM is proposed to overcome the shortcomings of the traditional ASM and AAM. ASM is used to locate the outer points, and the AAM is used to locate the inner points, which significantly improves the accuracy. Besides,the 2D contour is used on some key feature points to improve the accuracy of locating. The position information of face, eyes, nose and mouth is used to initialize the average face shape. The multi-scaled ASM model is used to accelerate the convergence rate. Experimental results show that our proposed method can achieve higher accuracy than the traditional ASM,AAM and the improved IASM, PAAM method.

关 键 词：人脸形状特征点定位 主动形状模型(ASM) 主动表观模(AAM) 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]