摘要： 提出一种用于图像识别的有监督直接局部保持投影算法，该算法结合样本类别信息，通过同时对角化的方法求解局部保持投影问题，避免矩阵的奇异性。在ORL人脸库上的测试结果表明，该算法的识别率高于PCA, PCA+LPP等方法。

关键词: 局部保持投影, 图像识别, 主成分分析

Abstract: This paper proposes a supervised direct locality preserving projection algorithm for image recognition. This algorithm solves locality preserving projection problem via simultaneous diagonallization combined with samples’ information, and can avoid the singularity of the matrices. Experimental results on the ORL face database show that this algorithm is more powerful than PCA and PCA+LPP for face recognition.

Key words: locality preserving projection, image recognition, Principal Component Analysis(PCA)

中图分类号: 

• TN911.73