基于局部线性投影的三维旋转人脸识别

Rotated Face Recognition based on Projective Locally Linear Embedding

本文在基于三维形态学模型的人脸识别研究中,运用局部线性投影算法(ProjectiveLLE,PLLE)取代了普遍使用的经典降维方法PCA,在高维空间中减小了降维过程中的线性失真。有效提高了所构建的三维形态学模型的质量,经实验证明该算法对于提高拟合的精度和大角度旋转人脸的识别率都取得了比较理想的效果。

In this paper, the Projective Locally Linear Embedding is adopted in the research of rotated face recognition based on 3D Morphable Model, which replaced the traditional dimension reduction method, PCA. Results of the experiment showed, it efficiently kept the real dimensionality estimation and either the fitting result or the recognition rate is improved after applying this algorithm.