【作者】 徐俊；

【导师】 黄贤武；

【作者基本信息】 苏州大学 ， 通信与信息系统， 2006， 硕士

【摘要】 随着网络及社会信息化的发展,准确的身份识别方法成为一个必须解决的问题。基于生物特征的身份识别方法以人体特征为身份识别依据,相对于传统身份识别方法具有更好的安全性、可靠性和有效性。人脸和指纹识别是最为常用的生物特征身份识别方法。人脸识别系统具有友好、设备通用的优点,但其识别率较低;指纹识别识别率较高,但图像采集需要专用设备,而且指纹识别速度较慢,尤其是对于大型数据库。这些单一的生物特征身份识别系统在实际应用中各有局限性,往往只能够满足一部分要求。因此,基于多种生物特征的身份识别技术正成为新的研究热点。本文对多种生物特征综合识别系统的特征提取算法进行研究,即人脸和指纹综合识别系统。首先对人脸图像进行预处理、定位人眼位置,进而根据人眼位置进行人脸图像的几何归一化,对归一化后的人脸图像本文提出了ICA与PCA相结合的方法提取人脸图像的特征:首先,利用主元分析法(PCA)对人脸图像数据进行降维,然后采用独立分量分析(ICA)算法提取人脸整体特征,从而得到较准确的人脸特征。指纹特征提取采用基于脊线跟踪的从灰度图像中直接提取特征的算法,避免了传统算法中因二值化、细化等繁琐的中间步骤而引入伪特征点的问题,并对算法中的指纹图像增强、方向值计算、脊线终止判断条件等进行了改进。实验表明,本文算法得到的人脸特征更有利于人脸分类且计算量大大减小,指纹特征更为准确且提高了特征提取速度,同时有利于提高综合识别系统的准确度和速度。更多还原

【Abstract】 As the development of Internet and information-based technologies, recognize a person’s identity accurately is a necessary problem. Technologies of identity recognition based on biometrics such as people’s characters have more security, reliability and validity compared with the traditional identity recognition method. Face recognition system is friendly and has the currency device, but the recognizing rate is relatively low. The recognizing rate of fingerprint recognition is relatively high, but the device for catching the fingerprint image is special, and the working speed is slow especially for the big data-base. Identity recognition system based on single biometrics has disadvantage in practice, and just satisfies some facts, so technologies of identity recognition based on multi-biometrics that combine kinds of human characters have been got a certain extent research.This dissertation research on the feather extracting algorithm of the technologies of identity recognition based on multi-biometrics which is face recognition combined with fingerprint recognition. First pre-process the face image and localize the eyes, then normalize the face image based on the location of eyes, and extract the face feather using the independent component analysis algorithm combined with the principal component analysis algorithm to reduce the face image dimension; Attempt a direct gray scale minutiae detection approach based on ridgeline following to extract fingerprint feather, it will overcome the shortcoming of traditional method which would bring in false feathers during the binarization and thinning, and make some improvement such as the fingerprint image enhancement, direction field, stop criteria of ending point. Experiments show that the face feather and the fingerprint feather is more accurate, more suitable for classify, and decreases the complexity of the algorithm, and improve the accuracy and the operating speed of the multi-biometrics system.更多还原