基于非接触采集下的鲁棒掌纹识别  

Robust palmprint recognition based on contactless collection

作　　者：桑海峰[1] 马月施[1] 黄静[1] 

机构地区：[1]沈阳工业大学信息科学与工程学院,沈阳110870

出　　处：《沈阳工业大学学报》2013年第2期187-193,共7页Journal of Shenyang University of Technology

基　　金：国家自然科学基金资助项目(60972123);辽宁省教育厅2012年度高等学校科研基金资助项目(L2012034)

摘　　要：为了扩大掌纹识别的应用范围,使非接触式掌纹识别系统在复杂背景和光照变化的条件下都能使用,提出了一种适用于非接触采集下的鲁棒掌纹识别算法.该算法使用肤色阈值的方法把手掌从复杂的背景中分割出来,采用一种新颖的指根点检测算法找到指根点,用指根点来定位掌纹ROI,使用对光照具有良好鲁棒性的LBP算法提取掌纹特征,并用Chi平方统计法进行分类.实验使用的数据库是在不同光照和复杂背景下采集的,当等错率为3.098 6%时,正确识别率达到97.833 7%.实验结果表明,该方法能够扩大掌纹识别的应用范围,使非接触掌纹识别系统在复杂背景和可变光照的条件下达到较理想的识别效果.In order to expand the application field of palmprint recognition and ensure the application of contactless palmprint recognition system under the conditions of both complex background and illumination variation, a robust palmprint recognition algorithm suitable for the contactless collection condition was proposed. The palm was segmented from the complex background with a skin-color thresholding method. Then a novel finger base point detection algorithm was used to find out the finger base points which could be adopted to locate the region of invest （ROI） of palmprint. The local binary pattern （LBP） algorithm exhibiting good robustness to illumination was applied to extract the features of palmprint. The classification was performed with Chi square statistics method. The database used in the experiments was collected under different illumination and complex background. When the equal error rate is 3.098 6%, the correct identification rate reaches 97. 833 7%. The experimental results prove that the proposed method can expand the application field of palmprint recognition, and make the contactless palmprint recognition system achieve the ideal effect under the conditions of complex background and illumination variation.

关 键 词：非接触式 掌纹识别 光照变化 复杂背景 肤色建模 指根点检测 局部二值模式 Chi平方统计 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]