【作者】 黄静；

【导师】 苑玮琦；

【作者基本信息】 沈阳工业大学 ， 计算机应用技术， 2007， 硕士

【摘要】 生物识别技术是利用人体所固有的生物特征来进行自动身份识别的技术。人体生物特征具有普遍性、唯一性和稳定性等特点,并且不会被遗忘,也较难被模仿或伪造。与传统的身份识别方法相比,生物识别技术更能适应现代信息社会对人体身份验证安全性和方便性的要求。基于掌纹的人体生物特征识别技术是近年来兴起的一个新的研究方向。掌纹含有丰富的信息,不像指纹一样易于磨损;设备成本要低于虹膜识别;更容易被用户接受;其识别精度要高于人脸识别;稳定性高于签名识别。由于上述原因,掌纹识别很快便成为生物识别技术领域中的一个研究热点。但目前国内外对掌纹的研究还远没有指纹、人脸、虹膜识别的研究那样广泛和深入。一个掌纹识别系统包括图像预处理、特征提取和分类器设计等几个模块。本论文针对掌纹的具体特点,对掌纹识别系统中预处理与特征提取的两个模块的算法进行了深入的研究。本文提出了一种掌纹图像预处理算法,使用该算法对采集到的掌纹图像进行定位,对掌纹图像定位的正确率达到了98.8%。在正确定位的基础上,本文使用主成份分析(PCA)方法提取掌纹特征,这种特征能反映图像本身所具有的内在特性。通过计算出多幅掌纹图像所构成的协方差矩阵的特征向量,来定义掌纹图像的投影子空间。为了克服主成份分析方法中计算量大的缺点,提出了一种新的掌纹特征提取方法,即基于小波分解与主成份分析的掌纹特征提取方法,其目的在于在不降低识别率的情况下,利用小波分解的多分辨率分析的特征,通过降低掌纹图像分辨率来提高掌纹特征提取速度。基于小波分解与主成份分析方法经过实验证明,本方法在识别率和计算复杂性方面都有令人满意的结果。更多还原

【Abstract】 Technology of biometrics uses human body characteristics for automated personal authentication. Because biometric features are universal, unique, stable, unforgettable and hard to imitate and forge. In contrast to the traditional methods, biometrics is now becoming a new way to meet security and convenience demand for personal identification.Personal authentication based on palmprint is a new branch of this area in the coming years. Palmprints contain more distinctive information than fingerprints, have less instruments cost than iris, obtain higher accuracy than face recognition, are easier to be accepted by clients, and are also more stable than signatures. All these characteristics make palmprint recognition as a hotspot soon after its appearance in the last few years. However, limited work has been reported on palmprint identification and verification at present.The design of a palmprint recognition system involves image data collection, image preprocessing, feature extraction, feature matching. This paper concentrates on the deeply research of the image preprocessing and feature extraction in palmprint identification system. It proposes a series of efficient palmprint preprocessing, feature extraction, feature matching algorithms, which can reach a 98.8% recognition rate. On the basis of proper localization, principal components analysis (PCA) can depict the distribution of main palmprint features, and define the projection subspace of palmprint by calculating the eigenvectors of the covariance matrix of palmprint images. In order to overcoming the deficiency of lage scale of calculation, a new method for feature extraction of palmprint is proposed. This method combines wavelet decomposing with PCA, which improves upon feature extraction speed of palmprint under without reducing recognition rate. It gets a satisfied experimental result showing the improvement in recognition rate and computational complexity based on analysis method of wavelet decomposition and PCA.更多还原