摘要
人脸识别是计算机视觉、模式识别中的一个研究热点和难点。针对墨镜和口罩这两个属性对人脸图像进行分类,系统地研究了不同的人脸分类算法的性能,其中包括主分量分析(PCA)、线性判别分析(LDA)、相关系数(Correlation)、支持向量机(SVM)、Adaboost算法,给出了在OMRON人脸库上的实验对比结果。实验表明,降低特征向量的维数,可以大大的减少分类时间的开销,同时不会明显降低分类器的性能。另外,本文提出了一种SVM和Adaboost相结合的方法,针对每种属性,选择和训练最优的人脸特征,取得了较为理想的结果。
Face recognition is one of the most active and challenging problems in machine vision and pattern recognition.Based on two facial attributes–sunglass and mask,some face classification algorithms,such as Principal Component Analysis(PCA),Linear Discriminant Analysis(LDA),Correlation Coefficient(Correlation),Support Vector Machine(SVM),and Adaboost are summarized in this paper.By using the OMRON face dataset for each method,the experimental results show that reducing feature dimensions could decrease the classification time largely,with very little harm to classification performance.In addition,a classification approach which exploits the strong structure of faces to select and train on the optimal set of features for each attribute is proposed in this artical,and the result seemed to be ideal.
出处
《微型电脑应用》
2010年第4期60-62,69,共3页
Microcomputer Applications
