摘要
针对人脸识别中由于环境光照、人脸姿态和表情等变化带来的识别困难,提出一种基于主动外观模型(Active Appearance Model,简称AAM)提取几何特征的人脸识别算法。本算法利用主动外观模型对人脸图像特征点进行准确的提取和定位;根据提取的特征点构造人脸几何特征向量并进行归一化,并根据几何特征向量之间的相似度简单快速地对人脸图像进行自动分类和识别。实验表明,本算法在主动外观模型基础上选取的几何特征向量具有尺寸、旋转和位移不变性,利用这些几何特征向量对人脸图像进行识别,可以有效避免人脸识别中常见的光照、姿态和表情影响并提高识别效率。
Due to the difficulty of face recognition caused by ambient light, face pose and expression change, a face recognition algorithm based on geometric characteristics extracted by Active Appearance Model （AAM） was proposed. This algorithm first made use of AAM to precisely extract and locate a set of feature pointsfrom the face images. Then, geometric characteristic vectors of faces were constructed and normalized. Finally, the face images were classified and recognized rapidly according to the similarity of characteristic vectors in different images. Experimental results demonstrate that the geometric characteristic vectors built on the basis of AAM have the strength of scalar, rotational and translational invariance. Using these vectors, the proposed algorithm can not only keep a good robustness to the variations of illumination, pose and face expression, but also improve recognition efficiency.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2013年第10期2374-2380,共7页
Journal of System Simulation
基金
国家自然科学基金(60903064)
山东省自然科学基金(ZR2011FQ003)
关键词
AAM
特征点
几何特征向量
人脸识别
AAM （Active Appearance Model）
feature points
geometric characteristic vector
face recognition
