基于改进AAM的人脸特征点检测  

Face Feature Point Detection Based on Improved AAM

作　　者：王磊[1] 刘辉[1] 余四洋[1] 毕学霞[1] 曾令军[1] 

机构地区：[1]昆明理工大学信息与自动化学院,云南昆明650051

出　　处：《贵州大学学报（自然科学版）》2012年第1期83-87,共5页Journal of Guizhou University:Natural Sciences

摘　　要：为了有效的解决传统AAM算法对光照的鲁棒性差,本文利用Gabor滤波器进行滤波处理,再通过LBP算法进行降维处理。针对该算法拟合结果受初始条件影响很大等因素,本算法利用Adaboost算法对人脸的旋转角度进行预估计,在加入了全局旋转平移因子,在提高了拟合速度的同时也取得了很好的拟合结果。In order to effectively solve the poor AAM robustness of the traditional algorithm of light,Gabor filters filter was used again through the LBP algorithm dimension reduction processing.According to this algorithm fitting results from the initial condition such as the influence factors being great,the algorithm used Adaboost algorithm of face of rotating Angle to estimate,after joining the global increase in the fitting of speed,very good fitting results were achieved.

关 键 词：AAM ASM ADABOOST 局部二元模式 人脸识别 特征点检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]