基于Canny-AAM的人脸特征定位算法  

Facial Feature Point Localization based on Canny-AAM

作　　者：颜丽 谷学静[2] Yan Li;Gu Xuejing(School of Information and Computer Engineering, PingXiang University, Pingxiang 337000;College of Electrical Engineering, Hebei United University, Tangshan 063009, China)

机构地区：[1]萍乡学院信息与计算机工程学院,江西萍乡337000 [2]河北联合大学电气工程学院,河北唐山063009

出　　处：《湖南科技大学学报（自然科学版）》2018年第2期66-71,共6页Journal of Hunan University of Science And Technology：Natural Science Edition

基　　金：国家自然科学基金资助项目(61170117);河北省高等学校科学研究计划重点资助项目(ZD2014074);河北联合大学博士科研启动项目资助(25003099)

摘　　要：主动表观模型（AAM）是经典的特征点检测方法,该模型对外轮廓的定位不是十分准确,且计算复杂度高,迭代次数大,容易陷入局部最优解,很难满足多姿态人脸特征点检测的要求.针对上述问题,提出了一种基于Canny边缘检测的AAM人脸特征定位方法.在使用AAM之前先利用Canny算法进行边缘检测,滤去非人脸轮廓部分像素点的纹理信息,使特征点极大限度的定位到人脸轮廓部分.实验结果表明：该方法减少了迭代次数,降低了计算的复杂度,提高了匹配的准确率,时间消耗降低了27.7%.Active appearance model （AAM） is a classic feature point detection method, this method positioning external contour is not very accurate, the calculation complexity is high, the number of iterations is big, easy to fall into local optimal solution, it is difficult to meet the requirements of more attitude face feature point detection. Aiming at this problem, a AAM facial location algorithm based on Canny edge detection was proposed. Before using AAM, Canny algorithm was used to detect the edge of image and to eliminate the texture information of non-face pixel, to make the feature points locate the face contour ultimately. Experiments show that the method reduce the number of iterations and the calculation complexity greatly. The proposed method reduced the average position error greatly and the time consumption is decreased by about 27.7%.

关 键 词：主动表观模型 CANNY边缘检测 人脸定位 人脸识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]