摘要
研究了基于动态模板与活动轮廓的人脸正面自动模式识别方法．用基于先验知识的积分投影提取人脸特征大致位置，给出针对人脸具体几何特征的动态模板（Ｄｅｆｏｒｍ ａｂｌｅＴｅｍ ｐｌａｔｅ）与活动轮廓（Ａｃｔｉｖｅ Ｃｏｎｔｏｕｒ）的数学模型及改进的能量函数．该模型能提取出人脸特征的精确轮廓，将特征输入Ｂ－Ｐ网络进行训练．实验证明，系统识别率高，且对于光照、倾斜等具有一定的不敏感性．
An automatic front view face recognition system is studied:First,coarse facial features are extracted by knowledge based integral projection.Then,mathematical models and energy functions of deformable template and active contour,which can extract the fine contour of each facial component,are introduced.B P neural network techniques are employed to construct a classifier.Through experiment,this system is proved to have a high recognition accuracy,and work well regardless of the variations in scale,tilt and rotation of head.
出处
《四川大学学报（自然科学版）》
CAS
CSCD
北大核心
1999年第5期898-902,共5页
Journal of Sichuan University(Natural Science Edition)
基金
国家自然科学基金
关键词
人脸
正面自动识别
动态模板
活动轮廓
B-P网络
automatic front view face recognition
deformable template
active contour
feature extraction
B P neural network
