基于改进PCA的人脸识别算法

Face recognition algorithm based on improved PCA

针对传统主成分分析法在特征提取中出现的耗时过长、平均对待所有像素点等问题,提出一种双中心羽化加权双向PCA(Bidirectional WPCA,BD-WPCA)的算法。算法首先将训练人脸样本和测试人脸样本图片进行双中心羽化加权处理,以增加人脸主要器官在识别中所占的比重;再用双向PCA算法分别在行和列方向上降维并提取特征;最后用K近邻法匹配分类。实验结果表明,该算法在降低运算耗费时间的同时能获得较高的识别率,具有可行性。

Aiming at the problems of original PCA like long calculating time and average treats all pixels, a recognition algorithm based on Bidirectional-WPCA was proposed. Firstly, the training and testing sample was weighted by double-center eclosion function to increase proportion of major organs. Then, the dimension of samples was reduced on the row and column by using the Bidirectional PCA algorithm. Last, the eigen was matched by K-NN algorithm. The experiment indicates that the algorithm not only speeded up the rate of calculation, but also got a better accuracy, which confirms the feasibility of the algorithm.