【作者】 蔡润斌；

【导师】 胡德文；

【作者基本信息】 国防科学技术大学 ， 控制科学与工程， 2010， 硕士

【摘要】 掌纹和掌脉识别技术是最近几年才发展起来的新的生物特征识别技术,以其独特的优势得到了国内外生物特征识别领域研究人员的广泛关注。本文在国家“863”相关课题的支持下,围绕掌纹和掌脉图像信息融合,开展了下述的研究工作:1)利用掌脉和掌纹的成像原理,本文提出了利用JAI公司的AD080CL多光谱摄像机采集掌纹和掌脉图像。由于是利用单个摄像机采集所以可以不必进行掌纹和掌脉图像的配准。2)提出了对掌纹和掌脉图像的预处理流程,因为在采集红外图像的时候必然引入噪声。预处理中采用小波模极大值方法实现在图像去噪和增强,同时也对感兴趣区域的选取做了一定的研究。3)利用双树复数小波和对抗融合技术实现掌纹和掌脉图像的融合,同时对融合规则做了重点介绍。随后也介绍了离散小波变换和平移不变小波变换的融合技术,主要用于和前述两个算法做对比。4)采用熵、互信息量和算法耗时对几个融合方法做了初步的评价,证明了双树复数小波和对抗融合方法的优越性。最后利用排序测度方法对单一的掌纹、单一的掌脉、双树复数小波和对抗融合方法做特征提取,继而进行特征匹配,识别效果表明采用信息融合的方法在性能上明显优于单一特征的方法。更多还原

【Abstract】 Palmprint and palm vein recognition technologies developed in recent years are new biometric technologies. With their unique advantages, palm vein and palmprint recognition attract widespread attention of the domestic and international researchers in the field of biometric. Supported by the related‘863’projects of the state, this thesis focuses on the fusion of palmprint image and palm vein image. Our contributions include:1) Considering the imaging principle of palmprint and palm vein, we propose to use AD080CL multispectral camera of JAL to get palmprint and palm vein image. Because these images are from the same camera, registrations of images are not needed.2) Preprocessing of palmprint and palm vein image is proposed, the reason is that infrared imaging system inevitably have palm vein image with noise. In the process of preprocessing wavelet modulus maximum algorithm are used for image denoising and image enhancement,and we also do some research on how to get ROI of palmprint and palm vein image.3) Dual tree complex wavelet tansform (DTCWT) and opponent processing are proposed for the fusion of palmprint and palm vein image, and fusion rules are emphasized. Also discrete wavelet transform (DWT) and shift invariant discrete wavelet transform (SIDWT) are introduced which are used to compare with the formal two methods.4) Entropy, mutual information and elapsed time are calculated, the experimental results show that DTCWT and opponent processing have more advantages than others. Finally ordinal mesure are used for feature extraction in single palmprint, single palm vein, DTCWT and opponent processing system, and by the result of feature matching we can conclude that information fusion methods have excellent performance than single one.更多还原