基于改进的ASM方法的人脸特征点检测  

Face feature points detection based on refined ASM

作　　者：孙延鹏[1] 傅蓉[2] 

机构地区：[1]青岛大学,山东青岛266071 [2]上海中科院计算技术研究所,上海201203

出　　处：《计算机工程与应用》2008年第10期163-165,共3页Computer Engineering and Applications

基　　金：国家自然科学基金(the National Natural Science Foundation of China under Grant No.60303018)

摘　　要：针对传统的Active Shape Mode(lASM)方法的缺陷与不足,提出了几点改进的方法。传统的ASM方法利用人脸检测提供的初始位置进行特征搜索;在改进算法中,利用了瞳孔定位的方法进行初始化,并采取了边缘约束,自适应步长等改进方法,有效地改善了ASM方法的性能。实验结果表明,改进的ASM方法在准确性和鲁棒性上有较大提高。In view of the traditional Active Shape Model （ASM） method＇s flaw and insufficiency,the paper proposes several improvements methods.The traditional ASM method carries on the characteristic search using of the initial point which has provided using the person face examination;in the impl＇oved algorithm,the paper has used the method of pupil orientation on the initialization and adopted improvement methods of edge restraint,auto-adapted length of stride to effectively improved the ASM method performance.The experimental result indicates that,the improved ASM method has a bigger enhancement in the accuracy and robustness.

关 键 词：主动形状模型 瞳孔定位 边缘约束 自适应步长 人脸识别 

分 类 号：TP31[自动化与计算机技术—计算机软件与理论]