基于局部方向场散列的掌纹快速辨识方法

面向大规模人群身份辨识的掌纹识别系统有广阔的应用前景。为了加速系统的响应时间,提出了一种基于局部方向场散列的快速掌纹识别方法局部方向场散列(LOFP hashing)。该方法首先使用一种方向特征双编码的方式,解决了传统编码方式不稳定、精度差的缺点。然后提出一种特征窗口化的处理方式,即局部方向场模式,加速目标掌纹的检索。在香港理工大学大规模掌纹数据库Hong Kong PolyU large-scale database,中国科学院掌纹数据库CASIA database以及一个大规模模拟数据库上进行对比实验,结果表明,提出的方法在基本不损失精度甚至有所提高的前提下,可以显著加速辨识速度。因此,LOFP hashing可以有效提升面向大规模人群掌纹识别系统的性能。

LOCAL ORIENTATION FIELD HASHING-BASED FAST PALMPRINT IDENTIFICATION

Palmprint recognition system for large-scale crowd identification has broad application prospects. In order to speed up the identification process, a novel hashing based technique called LOFP Hashing was proposed. This method first perfermed a double-orientation encoding method solve the disadvantages of the unstable and poor accuracy of the traditional encoding method. Then proposed a window-based feature measurement to speed up the process of searching the target template, referred as local orientation field pattern. Experimental results on two real public database and a synthetic database show that, the proposed method can significantly speedup the state-of-the-art methods. Therefore, LOFP Hashing can effectively improve the performance of palmprint identification system for large-scale crowds.