【作者】 肖潇；

【导师】 刘国海；

【作者基本信息】 江苏大学 ， 控制理论与控制工程， 2010， 硕士

【摘要】 随着现代社会对信息安全要求的不断提高,利用生物特征进行快速而准确的身份识别越来越受到人们的重视。静脉识别技术是一种新兴的非接触式红外生物特征识别技术,它不但识别率高而且安全性高、使用方便、实现容易,正逐步成为当前热门的研究课题。手指静脉识别技术根据近红外光可以被血液强烈吸收而被其他人体组织散射的特性,利用每个人的手指静脉分布不同这一特征进行身份识别。本论文在收集和分析近年来国内外有关生物识别技术研究成果的基础上,对手指静脉识别的关键技术进行了研究。本文重点完成了对手指静脉图像的预处理、特征提取以及匹配识别,在PC机上采用MATLAB 7.0对所有算法进行了仿真实验和分析,最后设计了手指静脉身份识别系统,具体内容如下:首先,对采集到的手指静脉图像进行预处理。实现对图像的边缘定位、归一化、滤波以及直方图修正处理。有效地去除了原始图像中的各种噪声,增强了图像的清晰度,为后文准确地提取手指静脉特征打下了基础。其次,对手指静脉图像进行特征提取。主要完成静脉拓扑结构的特征提取,即骨架特征。在分析传统算法的基础上,针对已有算法耗时长、准确性不高等问题,提出了一种有效去除干扰的快速特征提取方法。采用多尺度形态学变换,沿图像边界扫描,检测4个方向上手指静脉横截面灰度值所形成的谷形域,避免了逐像素点比较的缺点,实验结果表明,提取出的静脉连续性和细节完整性更优,算法耗时更少。再次,对手指静脉图像匹配算法进行研究。分别实现了基于Hu不变矩、Tchebichef正交不变矩,采用最近邻域特征的匹配算法,和基于BP神经网络的匹配算法。针对不变矩匹配算法识别率不高的问题,在BP神经网络算法中,提取手指静脉拓扑结构的几何特征,并结合矩特征构成新的输入特征向量。实验证明,基于BP神经网络的匹配算法在识别速度和识别率上都取得了更好的效果。最后,完成对手指静脉身份识别系统的设计。主要通过红外光源、CMOS图像传感器实现图像采集,ADSP BF561处理器完成算法处理。更多还原

【Abstract】 With the development of the information security requirement of modern society, using biometric character to identify one’s identification quickly and exactly thrives. Vein Pattern Recognition is a new contactless biometric technology using IR. It not only offers high accuracy personal identification, but also offers high safety, usability, and can be implemented easily. So it becomes a hot spot stage by stage. According to the characteristic of the infrared light that while the infrared light is absorbed intensively by the blood, it is dispersed by other organs of the body. The finger vein recognition technology is carried on the body’s identification through the finger vein. Based on the latest extensive discourses and technology journals in this field, the dissertation is trying to make studies on finger vein recognition.The finger vein image pre-processing, minutiae extracting and matching is mainly studied in the dissertation. The simulation is complemented by MATLAB 7.0 on PC and the finger vein identification system is designed at last. The contents are as follows:Firstly, the finger vein image is dealt with the image pre-processing technologies. Pre-processing technologies includes edge location, normalization, filter and Histogram Modification. Its purpose is to remove noise of vein image and improving definition of vein image, which will be beneficial to following feature extraction.Secondly, the features are extracted from the finger vein image. The features of finger vein’s topology which means skeleton features are mainly extracted. Based on the analysis of traditional means of segmentation, due to the fact that the algorithm is time-consuming, with low accuracy rate, the dissertation proposes a novel algorithm for finger vein’s skeleton features extraction which can remove interference effectively and improve the speed rapidly. With multi-scale Morphological transform, the images will be scanned across the edges and the valley detection will be done from the four directions. So, the shortcomings of the methods which compare the intensities pixel by pixel are avoided. Experimental results have proved that the finger vein extracted by the proposed method has more precise details and better continuity. The running time is reduced too.Thirdly, the search on recognition algorithm of finger vein images is implemented. In this dissertation, we implement two recognition algorithms, that is, based on Hu moment invariants and Tchebichef orthogonal moment invariants, with the nearest neighbor feature, and based on BP neural network. Against the low identification rate of the former algorithm, geometry in topology and moment features are used as input vector in BP neural network. Experiments have proved that the matching method based on BP neural network achieved good results on both speed and recognition rate.Finally, the finger vein identification system is designed. The hardware includes infrared LED, CMOS image sensor which is used for image acquisition and ADSP BF561 which is used to realize algorithm.更多还原