仿生算法与主成分分析相融合的人脸识别方法  

Face Recognition Algorithm Integrating Bionic Algorithm with Principal Component Analysis

作　　者：张祥德[1] 张大为[1] 唐青松[1] 陆小军[1] 

机构地区：[1]东北大学理学院,辽宁沈阳110004

出　　处：《东北大学学报（自然科学版）》2009年第7期972-975,共4页Journal of Northeastern University(Natural Science)

基　　金：辽宁省科学技术基金资助项目(002010)

摘　　要：基于人脸特征提取问题可以转化为组合优化问题这一思路,提出了仿生算法与主成分分析相融合的人脸识别算法.该方法先通过主成分分析方法得到人脸特征子空间;然后在已有特征的基础上,分别通过遗传算法与离散粒子群算法进一步提取出可使识别正确率达到最高的人脸图像特征.在ORL人脸库上的实验结果表明:与传统的主成分分析相比,该方法不仅能进一步降低特征子空间的维数,从而提高识别速度,而且能获得更高的识别率.Considering that the face feature extraction problem can be transformed into a combinatorial optimization one, a new face recognition algorithm is proposed to integrate the bionic algorithm with principal comment analysis. In the new method, the feature subspace of face pictures is obtained through conventional principle component analysis. Then, the most accurate recognition is given from the subspace by use of the genetic algorithm and binm3r particle swarm algorithm separately. The experimental results based on ORL face database showed that the new algorithm is able to reduce the number of dimensions of feature space with highly accurate recognition rate.

关 键 词：遗传算法 粒子群优化算法 主成分分析 人脸识别 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]