随着社会信息化网络化的发展，信息安全变的越来越重要。传统的身份认证方式已经难以满足信息社会的需要，因此人们将目光投向了生物特征识别这个广阔的领域。掌纹识别作为一种可靠的生物特征识别技术，具有重要的理论研究价值和应用前景，吸引着国内外众多的研究人员。经过多年的研究，自动掌纹识别领域已经获得大量的研究成果。 然而与科研方面的如火如荼不同，自动掌纹识别在现实社会生活中并没有得到有效的应用。相对于已经被广泛使用的指纹识别等方式，掌纹识别仍然有许多问题有待解决。本文立足于掌纹识别在多模式识别方面的应用，在掌纹分类方面做出一些研究，希望使掌纹识别这个古老的身份认证技术在新的时代背景下能够获得广泛应用。 本文对以下三个方面做了研究： 一、通过对掌纹采集过程和掌纹图像特点的分析，本文提出了使用自商图像算法的掌纹图像预处理方法，该方法可以将图像中的掌纹信息映射到一个平面上，它是本文分类方法的基础。 二、本文提出一种基于快速匹配思想的掌纹分类方法，该方法从掌纹图像中提取纹线统计信息，并使用神经网络对它们进行快速匹配。本文使用香港理工大学（PolyU）掌纹数据库对该方法进行实验，实验结果表明这种分类方法可以在一定程度上减少匹配工作量。 三、提出一种基于全局特征的掌纹分类方法，该方法使用共生矩阵计算图像的纹理特征作为分类指数，同样在PolyU掌纹数据库上进行实验，实验结果表明该分类方法有一定的分类效果。; Recently，with the rapid development of information and internet throughout the world，information security shows its significant importance. Due to many drawbacks in traditional authentication methods, such as keys or passwords, now people turn to biometrics, which is a new research field with brilliant future. As one of the reliable recognition technologies, palmprint recognition has attracted extensive attention of researchers. After years research, there has been fruitful results in palmprint recognition fields. However, palmprint recognition has not been applied intensively in the real world. Compared to mature fingerprint recognition approaches, there are still a lot of unsolved problems. This paper focuses on the application of palmprint recognition algorithms in the multi-pattern recognition field. Some progresses have been made in the palmprint classification algorithm in order to obtain the wider application scope. The main contribution of this dissertation lies in that: First, by the analysis of the processing of palmprint capture and the characteristic of the palmprint image, a preprocessing method is proposed for the palmprint image using the self-quotient image algorithm. The proposed method projects the palmprint image onto a special plane, and it is the base of our palmprint classification algorithms. Second, a palmprint classification algorithm is suggested based on the fast features matching. This algorithm extracts the statistic information of palmlines from palmprint image, and neural networks have been used for feature matching. The experiment is done on the PolyU palmprint database, and the results manifest that the proposed method can reduce the matching work to a certain degree. Third, a palmprint classification algorithm is proposed based on the global characteristic of the palmprint. This method utilizes the co-occurrence matrix to compute the classification index, and this index is then used to do experiments on the PolyU palmprint database. The experimental results show that this method can reduce the matching work to 42% of the original.
修改评论