基于灰度曲面匹配的快速手掌静脉识别

Fast Palm Vein Identification Algorithm Based on Grayscale Surface Matching

为使掌脉识别系统在识别性能和识别时间上有一个较好的平衡,提出了一种基于灰度曲面匹配的快速手掌静脉识别算法。对手掌静脉图像提取感兴趣区域,将感兴趣区域等分为若干个子区域,计算每个子区域像素灰度平均值作为该子区域灰度值,以各子区域灰度值构建待匹配图像。匹配时对两个待匹配灰度曲面中的像素灰度做差,得到灰度差曲面,求出该灰度差曲面的方差,将此方差作为衡量两个掌脉特征曲面之间距离的依据,并据此判定两幅掌脉图像是否来自同一只手。应用自建掌脉图库进行实验分析,该算法选择子区域大小为8pixel×8pixel时的正确识别率达到97.94%,识别时间仅用0.163ms。实验结果表明,与传统掌脉识别算法相比,该算法在识别性能和识别时间上有一个较好的平衡。

In order to improve the recognition speed with effective recognition performance of palm vein identification system, a fast palm vein identification algorithm based on grayscale surface matching is proposed. The algorithm extracts region of interest （ROD of palm vein image firstly. Then, the ROI is equally divided into several sub-regions. The algorithm computes average value of the grayscale of every sub-region. These average values construct an image for matching. At the stage of matching, the algorithm computes the difference between two pixels from two matching images and gets the grayscale difference surface. It calculates the variance of the grayscale difference surface and considers this variance as the distance between two feature surfaces. At last, it decides whether these two images come from the same hand or not according to the variance. A self-built palm vein database is used in the experiment. The experimental result shows that the scheme with sub-region parameter of 8 pixel× 8 pixel reaches correct recognition rate （CRR） of 97.94%, with recognition time of only 0. 163 ms. Compared with the traditional palm vein recognition method, the proposed method increases recognition speed with effective recognition performance.