基于图像旋转变换的改进PCA与LVQ的人脸识别

Face Recognition Method with Improved PCA and LVQ Network Based on Image Rotation

为有效解决BP分类器训练时易震荡,易陷入局部极小值和人脸图像由于受拍摄角度,表情变化等因素影响而导致识别率低的问题,提出一种基于图像旋转变换的改进的主成分分析(PCA)与学习向量量化网络(LVQ)相结合的新算法。首先用辐射模板对非正面人脸进行标准化,然后将PCA和具有多方向,多尺度滤波特性的二维Gabor相结合进行降维,最后使用鲁棒性强,结构简单的LVQ网络进行分类识别。本文算法利用ORL人脸库进行仿真,证明了此方法的可行性。

In order to effectively solve the saituation that easy to shock when BP classifier training,fall into local minima and face images affected due to the shooting angle,facial expression changes and other factors which led to the problem of low recognition are proposed based on improved image rotation transformation of primary component analysis( PCA) and learning vector quantization new algorithm network( LVQ) combine. Firstly with radiation for non-frontal face template to standardize,then classify PCA and Gabor filter which has multi-directional,multi-scale two-dimensional characteristics combine to reduce the dimension. Finally,distingnish with the simple network structure robust and good classification LVQ network. The algorithm uses the ORL database simulation to demonstrate the effectiveness of this method.