手指静脉图像的概率分割方法研究  

Research on Finger-vein Image Segmentation Based on Probability Maps

作　　者：温梦娜 杨金锋 WEN Meng-na;YANG Jin-feng(College of Electronic Information and Automation, Civil Aviation University of China, Tianjin 300300, China;Tianjin Key Lab for Advanced Signal Processing, Civil Aviation University of China, Tianjin 300300, China)

机构地区：[1]中国民航大学电子信息与自动化学院,天津300300 [2]中国民航大学天津市智能信号与图像处理重点实验室,天津300300

出　　处：《小型微型计算机系统》2018年第7期1569-1573,共5页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61379102;U1433120;61502498)资助;中央高校基本科研业务费专项资金项目(3122017001)资助

摘　　要：由于光在生物组织内部会产生严重衰减,手指静脉特征的图像质量往往较差,这十分不利于实现静脉区域的稳定分割.为了较为精确地获取手指静脉网络,本文提出了一种新的基于卷积神经网络的手指静脉区域分割方法.首先,利用韦伯定律去除光照变化实现对手指静脉图像的增强.然后,通过自动粗略标注静脉区域与非静脉区域,获得带有标签的像素训练集.利用训练集,训练一个可产生像素属于静脉区域和非静脉区域概率的卷积神经网络模型.最后,利用概率图,通过概率运算实现对手指静脉图像的分割.实验结果表明,通过该方法能够得到较为理想的静脉网络.Due to severe light attenuation while transmitting in tissue,the quality of finger-vein images is often undesirable for implementing finger-vein region segmentation. In this paper,a new scheme based on CNN（ convolutional neural network） is proposed for finger-vein network segmentation. Firstly,the Weber law is used for image enhancement by removing illumination variation in fingervein images. Secondly,based on the enhanced images,the training pixel sets are constructed by automatically labeling venous regions and none-venous regions. Feeding the labeled pixel sets with venous and non-venous labels into the developed CNN structure,a CNN model is trained for generating probability maps of venous and non-venous regions. Using the generated probability maps,the fingervein images are finally segmented. The experimental results show that the finger-vein networks can be obtained desirably by the proposed method.

关 键 词：手指静脉图像 韦伯定律 图像分割 卷积神经网络 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]