基于特征参量空间的掌脉识别方法研究

Study on palm vein recognition method based on feature parameter space

非接触式采集是手掌静脉识别的主流,但其低约束性可能导致掌静脉存在平移、旋转和比例缩放,同时,手掌过度伸展可能导致部分静脉分布信息丢失,这些都可能影响掌脉识别结果。针对以上问题,提出一种新的掌静脉特征识别方法。首先,获取靠近指根部的手掌内切圆,利用内切圆内静脉交叉点和内切圆圆心定义相对半径及相邻夹角参数;其次,由相对半径和角度参数建立二维特征向量空间,将静脉交叉点转换成该特征向量空间内的一系列特征点;最后,计算特征点间的特征向量距离作为匹配点对的判断依据,根据匹配点对的比例得到匹配率。通过自建的掌静脉图库和CASIA图库对算法性能进行验证,分别得到等误率0.97%和4.98%。结果表明,该方法在手掌静脉产生比例缩放、旋转和平移后仍可获取较好的识别效果,同时具有特征点提取容错性。

Abstract：Contactless acquisition is the mainstream of palm vein recognition. However, the low-constraint of palm leads to the translation, rotation and scaling of palm vein image;hand hyperextension may result in vein blood ob- struction, which would cause the loss of local vein information ; and all of these problems may influence the palm vein recognition results. To solve these problems, a novel palm vein recognition method based on feature parameter space is proposed. Firstly, the inscribed circle of palm closest to finger edge is obtained. The relative radius and relative angle are defined by using the intersection points inside the inscribed circle and the inscribed circle center. Then, a two-di- mensional feature vector space is established from the relative radius and angle parameters. And the intersection points are converted to a series of feature points in the feature space. Finally, the eigenvector distance amongthe fea- ture points are calculated as the judgment basis of the feature point pairs. Matching rate can be obtained according to the proportion of the feature point pairs. The performance of the proposed method is verified with self-built palm vein image database and CASIA database,the obtained equal error rates （EER） are 0.97% and 4.98% ,respectively. Experimental results show that the method can obtain good recognition effect. The method has invariant to scaling, ro- tation and translation, and also has fault tolerance in feature point extraction.