基于三维点云匹配的手掌静脉识别  

Hand Vein Recognition Based on Three Dimensional Point Clouds Matching

作　　者：张祺深 周雅[1] 胡晓明[2] 王丹婷[1] 

机构地区：[1]北京理工大学光电学院,北京100081 [2]北京理工大学生命学院,北京100081

出　　处：《光学学报》2015年第1期266-276,共11页Acta Optica Sinica

基　　金：国家自然科学基金(30900385)

摘　　要：针对现有手掌静脉认证系统误拒率较高以及不支持大数据集匹配的问题,设计了基于透射式光源的双目视觉静脉三维点云重建装置,提出了基于三维点云匹配的手掌静脉认证算法。系统使用850 nm透射式发光二极管(LED)光源作为照明装置,由双目摄像机拍摄静脉视差图像进行三维重建。选择手掌静脉作为特征点描述其空间三维结构,提出了一种改进的内核相关性分析方法匹配三维点云。针对200组点云数据的实验结果验证了该方法的可行性和有效性,识别率达到了98%,误拒率2%,误识率0%,总特征维数约8000至12000维,高于尺度不变特征变换(SIFT),支持对大数据集的认证识别。In order to solve the problem of high false rejection rate and not supporting large data base registration in existing hand vein recognition system, a binocular stereoscopic vision device for hand vein three dimensional （3D） point can reconstruction is proposed, along with the hand vein 3D point cloud matching algorithm. The hand is lighted by an 850 nm light emitting diode （LED） light source, binocular images for 3D reconstruction are obtained by the stereo cameras. The hand vein＇s spatial structure is described by hand veins feature, an optimized kernel correlation analysis approach is proposed for 3D point cloud matching. Experimental results of 200 different point clouds data show the proposed system is feasible and effective, the recognition rate is 98%, false rejection rate is 2% and the false accept rate is 0%, the feature＇s dimension is ranged from 8000 to 12000, which is higher than that of scale invariant feature transform （SIFT）. The proposed system provides a possibility for large database recognition.

关 键 词：机器视觉 静脉三维匹配 三维重建 点云匹配 内核相关性分析 

分 类 号：O436[机械工程—光学工程]