摘要
人脸识别是生物特征识别技术最友好的身份识别方式,而三维人脸识别由于可解决二维人脸识别中存在的光照、姿态等局限,成为人脸识别的研究热点,但其特征维数过高是该领域的瓶颈,而维数约减是解决这一问题的关键。流形学习是一类非线性维数约减算法,LOGMAP是一种基于黎曼法坐标的流形学习算法,该算法可以在保持度量信息不变的情况下,把高维空间的数据映射到低维空间。构建了基于流形学习的三维人脸识别框架,并结合LOGMAP进行三维人脸深度图的识别。实验结果表明,该方法在三维深度图上可以得到良好的识别效果。
Face recognition is the no-touch authentication technology in biometrics. 3D face recognition becomes more popular because it can solve the problems of the effect from different illumination and pose in 2D face recognition. However, how to decrease the number of feature dimension is a key technology in 3D facial recognition. In this paper, LOGMAP manifold learning algorithm based on normal coordinates recently is proposed to establish the mapping relationship between the observed and the corresponding low-dimensional data. This paper built a 3D face recognition framework based on manifold learning, and applied LOGMAP theory into 3D facial depth image recognition. The experimental results demonstrate that this method can get good effect for recognition.
出处
《电子测量与仪器学报》
CSCD
2012年第2期138-143,共6页
Journal of Electronic Measurement and Instrumentation
基金
国家自然科学基金(61174170)
教育部博士点基金(2010111110005)资助项目
关键词
三维深度图
人脸识别框架
流形学习
黎曼法坐标
LOGMAP
3D facial depth map
face recognition framework
manifold learning
Riemannian normal coordinates
LOGMAP
