中文摘要：

针对手背静脉识别技术,提出了一种基于深度学习和多尺度编码组合的手背静脉识别算法.首先,利用下采样和小波分解获取多尺度下的手背静脉图像,然后采用中心对称的局部二值模式（CSLBP）提取图像的特征,再次对提取的特征采用深层模型—限制玻尔兹曼机（RBM）逐层训练,最后采用多尺度编码组合的方式进一步提高识别率.实验证明,本文所提出的方法较传统的PCA、LBP算法识别率更高.

英文摘要：

Based on the deep learning and multi-scale coding, this paper proposed a hand dorsal vein recognition algorithm. First, we acquired hand dorsal vein image of multiple scales by using sampling and wavelet decomposition. Second, we extracted image feature by center-symmetrical local binary patterns （CSLBP）. Third, we trained extracted features layer by layer again using deep model-Restricted Boltzmann Machine（RBM）. And finally we improved the recognition rate by adopting mode of multi-scale coding combination. The result shows that the recognition rate of the proposed method is higher than that of the traditional PCA／LBP algorithm.