主成分分析重建误差掌纹识别方法  

PCA Reconstruction Error Palmprint Recognition

作　　者：苑玮琦[1] 曲晓峰[1] 柯丽[1] 黄静[1] 

机构地区：[1]沈阳工业大学视觉检测技术研究所,辽宁沈阳110023

出　　处：《光学学报》2008年第10期1903-1909,共7页Acta Optica Sinica

基　　金：国家自然科学基金(60672078;60472088)资助项目

摘　　要：主成分分析(PCA)法在掌纹识别方面可以取得较好的效果。但是随着掌纹图像库的扩大,PCA转换矩阵训练时间迅速增长;注册新掌纹时,需要重新训练PCA转换矩阵,添加注册掌纹的代价随着掌纹库的增大迅速增加。如何能够在保持PCA识别效果的情况下提高使用的便捷性成为PCA广泛应用的主要障碍。提出了一种以PCA重建误差为分类依据的PCA重建误差掌纹识别方法。该方法与PCA法基于相同的原理,在采用最近邻分类器时可以取得与PCA法相等的性能;同时可以有效减少掌纹图像库的识别时间,可以以极少的代价扩展掌纹库。Principal component analysis （PCA） is very effective in palmprint recognition. However, with enlargement of the palmprint image library, the training time of the transforming matrix increases rapidly. When new palmprintis added in, the PCA transforming matrix should be retrained, so the cost of adding new palmprint to the libarary increases rapidly with the enlargement of the palmprint image libarary. It has been a major obstacle to the extensive use of PCA that how to make PCA easy to use, meanwhile the recognition performance is maintained. A PCA reconstruction error palmprint recognition method in which the PCA reconstruction error is taken for classification basis is proposed. The proposed method bases on the same theorem with PCA, so it can achieve the same performance with PCA when using the nearest neighbour classifier. The recognition time can be decreased sharply, and it is very easy to enlarge the palmprint image library with proposed method.

关 键 词：模式识别 掌纹识别 主成分分析 重建误差 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]