中文摘要：

研究了一种适用于非接触式图像采集的手掌静脉识别的子空间方法，解决了传统接触式采集容易传染疾病，非接触采集使同类图像差别增大导致识别性能不佳的问题。先采用分块算法对图像进行快速降维，再用偏最小二乘算法提取掌脉图像中灰度值变异大，且类别信息相关性最大的若干方向组成分类子空间，然后依据图像在此空间中的位置进行分类识别。应用自建掌脉图库和中科院自动化研究所图库进行实验分析，实验结果表明：与传统掌脉识别方法相比，该方法能有效地提高正确识别率，降低误拒率。两个图库中，该算法选择分块大小为4×4时的正确识别率分别达到99．98％，99．34％；误识率分别达到0．02％，0．66％；误拒率分别达到0．13％，0．60％；识别时间分别在0．03s，0．04s之内。适用于安防、考勤等场合，具有实用价值。

英文摘要：

This paper studies a palm vein recognition subspace method suitable for non-contact image acquisition. The problems that traditional contact palm vein image acquisition is subject to infecting diseases, and non-contact palm vein image acquisition increases the dissimilarity in intra-class and leads to low recognition ratio are solved. The image block algorithm is used to rapidly reduce the dimension firstly. Then ,the partial least squares algorithm is used to extract some directions ,in which the gray level in the palm vein image varies obviously and classical information correlation is maximal, to compose the classification subspace. The position coordinates of the image in this subspace are used to carry out classification and recognition. A self-built palm vein image database and the image database developed by Chinese Academy of Sciences Institute of Automation （CASIA） were used to carry out experimental analysis. The experiment results show that compared with traditional palm vein recognition method ,the proposed method effectively increases the Correct Recognition Rate （CRR） and decreases the false rejection rate （FRR）. In these two image databases,using this method and selecting the block size of 4 ×4 ,the CRR reaches 99.98% and 99.34% ;the false accept rate （FAR） reaches 0. 02% and 0.66% ; the FRR reaches 0.13% and 0.60% ,respectively;and the recognition time is within 0.03s and 0. 04s,respectively. The system is applicable in security and attendance management occasions ,and has practical value.