为了提高人脸识别方法对光照、姿态等外部因素的鲁棒性，本文在二维局部保持投影（2DLPP）算法的基础上进行改进，提出的一种双向2DLPP算法。与2DLPP算法不同的是，在求得行方向投影矩阵后，再求列方向的投影矩阵，得到图像的双向特征矩阵，以达到将样本降维的目的。实验结果表明，该方法具有较高的识别率对光照和姿态的变化具有一定的鲁棒性。

To improve the robustness for  the face recognition method of illumination,and pose external factors,this paper modifies the twodimensional locality preserving projections (2DLPP) and gives a bidirectional 2DLPP (Bidirectional 2DLPP) algorithm. The difference with 2DLPP is that the bidirectional feature matrix is obtained by seeking the projection matrix in the column direction after the projection matrix in the row direction,thus the purpose of the dimensionality reduction is achieved. The experimental results show that this method has some robustness and higher recognition rate for the variation of illumination and pose.