人体手指静脉图像识别算法研究

发布时间：2018-05-07 20:10

本文选题：指静脉识别 + 双向投影定位　；参考：《东华理工大学》2016年硕士论文

【摘要】：21世纪信息安全问题倍受社会关注,而传统的身份认证技术(如个人身份证、驾驶证等)已远不能满足信息安全需求,于是生物识别技术应运而生。目前,使用最为广泛的生物识别技术为指纹识别。但随着科技的不断发展,伪造人体指纹亦不是难事,甚至可以花几元钱就可以在淘宝网上购买到硅胶指纹膜,来自行克隆指纹。而这种克隆指纹还能够以假乱真,轻松骗过指纹识别仪,故存在重大的安全隐患。所以鉴于指纹识别自身的缺点,其发展已遇到了瓶颈。而人体指静脉因其具有唯一性(医学界已经验证)、生长在肉里、不易伪造等诸多优点,现已成为众多学者研究的新宠。本文研究内容主要围绕以下几点展开:(1)本文介绍了手指静脉图像采集原理:即是由于人体血液中的失氧血红蛋白能够吸收某一特定波长的近红外光线,也就是手指静脉能在近红外光下成像的方法。在对图像进行预处理阶段时,提出了一种基于ROI(感兴趣区域)的手指静脉图像定位的改进算法。经过验证,该算法解决了手指的平面转动问题定位问题,并获得了较满意的定位效果。(2)本文针对几种典型的手指静脉图像分割算法:均值法、Niblack法、OTSU法及迭代法进行了模拟仿真实验。实验结果表明:这些算法均不能提取出高质量的静脉纹路,分割结果也不是十分理想,且存在大量噪声和静脉特征。故本文提出了一种基于最大曲率与谷形分割相结合的改进算法,得到了很好的分割效果。然后对分割后的指静脉图像进行滤波、去噪与细化处理。(3)本文提取了指静脉骨架化图像中的两种关键性节点:交叉点和端点。然后利用修改的Hausdroff距离(MHD)来计算它们的相似度。实验表明:同时利用两类细节点作为特征点进行识别,能够有效地应用于个人身份认证。本文还对多模态生物识别系统的优势、层次、融合算法的结构以及模式源的选取作了简要介绍。最后简单分析了指纹与静脉的多模态生物识别系统的构建方法。
[Abstract]:In the 21st century, the problem of information security has attracted the attention of the society, but the traditional identity authentication technology (such as personal identity card, driver's license, etc.) has been far from meeting the information security needs, so biometric identification technology has emerged as the times require. At present, the most widely used biometric technology is fingerprint identification. But with the development of science and technology, it is not difficult to forge human fingerprint. It can even cost a few yuan to buy silica gel fingerprint film on Taobao to clone its own fingerprint. And this kind of clone fingerprint can also be false, easy to deceive fingerprint recognition instrument, so there are major security risks. Therefore, in view of the shortcomings of fingerprint recognition, its development has encountered a bottleneck. The human digital vein has become a new favorite of many scholars because of its uniqueness (the medical field has proved that it is growing in meat, it is not easy to forge, and so on. This paper mainly focuses on the following points: this paper introduces the principle of finger vein image acquisition: that is, the oxygenated hemoglobin in human blood can absorb a certain wavelength of near-infrared light. This is the way in which the finger veins can be imaged in near-infrared light. In the stage of image preprocessing, an improved algorithm of finger vein image location based on ROI (region of interest) is proposed. It is proved that the algorithm solves the problem of the plane rotation of fingers. In this paper, several typical finger vein image segmentation algorithms: the mean value method / Niblack / Otsu method and the iterative method are simulated and simulated. The experimental results show that these algorithms can not extract high quality veins, and the segmentation results are not ideal, and there are a lot of noise and vein features. Therefore, an improved algorithm based on the combination of maximum curvature and valley segmentation is proposed, and a good segmentation effect is obtained. Then the segmented digital vein image is filtered, denoised and thinned. (3) in this paper, two key nodes in the digital vein skeleton image are extracted: intersection point and end point. Then the modified Hausdroff distance is used to calculate their similarity. Experiments show that two kinds of detail points are used as feature points to identify and can be effectively applied to personal identity authentication. This paper also briefly introduces the advantages, levels, structure of fusion algorithm and the selection of pattern source of multimodal biometric recognition system. Finally, the method of constructing multimodal biometric identification system of fingerprint and vein is simply analyzed.
【学位授予单位】：东华理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41

【相似文献】