面相识别之贝叶斯方法的探析

Analysis of the Bayesian Method for Face Recognition

人脸识别问题是应用数学、模式识别领域的研究热点之一.在公安、司法监控等相关领域具有较大的应用前景.自动人脸识别系统包含人脸检测、特征提取、识别3个部分,本文讨论的是不含人脸检测的半自动人脸识别系统,贝叶斯方法是模式识别中的最优方法之一.本文在主成分分析法基础上重点讨论了贝叶斯人脸识别算法以及模型分布问题.在算法的推导中,将最大似然准则与最大后验准则作了比较并对冗繁的计算进行了改进,实验结果表明改进后的最大似然算法更有利于提高人脸识别的运算效率,而且识别率也比主成分分析方法有一定的提高.

Face recognition is one of the hot spots in application of mathematics and pattern recogni- tion. It has a large application prospect in the public security, judicial monitoring and other related ar- eas. This paper mainly discusses the Bayesian face recognition algorithm and the model distribution problem. In the algorithm derivation, the maximum likelihood criterion and the maximum a posteriori criterion are compared and the calculation was improved. The experimental results show that the im- proved maximum likelihood algorithm is more effective for improving the operation efficiency of face recognition.