基于手背静脉图像多特征稀疏表示的身份识别

Identity recognition based on multi feature sparse representation of dorsal hand vein image

为了准确地对人的身份进行识别,提出了一种对采集静脉图像的全局特征和局部特征进行稀疏表示的识别算法。该算法首先应确定静脉样本库中所有的静脉对象,并在不同光强下对每一手背静脉进行采集,此外将采集图像进行适当压缩与旋转,并将变换后的所有图像作为库中描述该静脉对象的样本;其次,分别提取该静脉对象所有样本的全局特征与局部特征,并通过求解每一特征系数向量的最小1范数,对未知静脉图像的全局与局部特征进行稀疏表示;最后,融合稀疏表示结果,完成静脉识别的过程。通过在3种光强下对200个人的手背静脉进行采集,并经过图像压缩与旋转调整后建立实验所需的静脉样本数据库,识别实验表明该识别方法正确识别率达到98%以上,并且对于采集时出现多种不合作因素具有较好的鲁棒性,同时具有较好的实用价值。

In order to recognize individual identity accurately,the global and local features of acquired dorsal hand vein image are extracted,and a vein recognition algorithm based on the sparse representation of those features is proposed.In the proposed algorithm all vein objects in vein sample database should be determined firstly,and every dorsal hand vein is acquired in different light intensities.Moreover,all the acquired vein images are compressed and rotated properly,and all the transformed images could be taken as the samples of the vein objects.Secondly,the global and local features of all the vein object samples are extracted respectively,and the global and local features of unknown vein image are represented sparsely through solving the minimum 1-norms of eigenvector coefficients.Finally,the results of sparse representation are fused,and the vein recognition process is completed.In the experiment,the vein sample database was established through acquiring 200 vein objects in three light intensities and shrinking and rotating the images;recognition experiment shows that the recognition rate of the proposed algorithm has exceeded 98%,and the algorithm has good robustness for multiple uncooperative factors and good practical value.