基于SIFT和Gabor变换的两类掌纹识别算法研究

发布时间：2018-06-04 06:46

本文选题：生物特征识别 + 掌纹识别　；参考：《重庆师范大学》2017年硕士论文

【摘要】：随着电子技术的快速成长,身份认证技术至关重要。掌纹识别技术,作为一种新兴的生物特征识别技术。相对其它的,具有难以伪装,比较稳定、不具侵犯性以及难以隐藏和非接触等特性。近年来掌纹识别技术在门禁、核电站、银行系统还有快捷支付中的应用,说明关于掌纹识别技术的研究是一个比较前景和具有现实意义的研究课题。掌纹识别的关键就是在于提取掌纹的有效特征,并根据提取到的掌纹特征设计相应的分类器进行分类识别。本文提出的,基于SIFT和Gabor小波变换的两类识别方法。主要从图像区域分割,特征提取以及分类识别三个方面进行了研究。本文主要完成的研究工作如下。1、掌纹图像的轮廓是通过顺序统计滤波、二值化和形态学的处理得到的,然后将检测到的角点中曲率最大的两个点作为定位点来建立坐标,从而提取感兴趣区域(ROI)。这种方法能有效分割图像。2、设计了Gabor变换和KPCA结合的一种掌纹识别算法。首先提取掌纹图像的四维四向变换,提取掌纹的纹理特征,将16组小波特征归一化,然后,执行核主成分分析(KPCA),并提取主成分,然后将主成分合并得到最终的向量特征,利用最近邻分类器进行识别。相对于单一的Gabor和KPCA算法,该算法的识别率更高。3、设计了SIFT特征匹配的一种掌纹识别算法。该算法利用待识别SIFT掌纹特征关键点,与已有SIFT掌纹特征关键点的匹配统计特性,计算累计匹配点数量的数值和阈值来进行身份识别。在matlab环境下仿真,结果表明,算法的识别率很高。
[Abstract]:With the rapid development of electronic technology, identity authentication technology is very important. Palmprint recognition technology, as a new biometric recognition technology. Compared with others, it is difficult to disguise, stable, non-invasive, difficult to hide and non-contact. In recent years, the application of palmprint recognition technology in entrance guard, nuclear power station, banking system and fast payment system shows that the research on palmprint recognition technology is a relatively promising and practical research topic. The key of palmprint recognition is to extract the effective features of palmprint and design the corresponding classifier according to the extracted palmprint features. In this paper, two kinds of recognition methods based on SIFT and Gabor wavelet transform are proposed. In this paper, three aspects of image region segmentation, feature extraction and classification recognition are studied. The main work of this paper is as follows: 1. The contour of palmprint image is obtained by sequential statistical filtering, binarization and morphological processing, and then the two points with the largest curvature in the detected corner are taken as the positioning points to establish the coordinates. Thus the region of interest was extracted. This method can effectively segment image. 2. A palmprint recognition algorithm combining Gabor transform and KPCA is designed. First of all, we extract the four-dimensional four-direction transformation of palmprint image, extract the texture feature of palmprint, normalize the 16 groups of wavelet features, then perform kernel principal component analysis (KPCAA), extract the principal component, then merge the principal component to get the final vector feature. The nearest neighbor classifier is used for recognition. Compared with the single Gabor and KPCA algorithms, the recognition rate of this algorithm is higher. 3. A palmprint recognition algorithm for SIFT feature matching is designed. The algorithm uses the key points of SIFT palmprint feature to be identified and matches the statistical characteristics of the existing SIFT palmprint feature points to calculate the value and threshold of the cumulative number of matching points to carry out identification. Simulation results under matlab environment show that the recognition rate of the algorithm is very high.
【学位授予单位】：重庆师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】