基于模块化的多算法的虹膜实验平台的设计与实现

发布时间：2018-02-27 21:27

本文关键词： 虹膜识别实验平台算法模块实验流图像定位特征提取　出处：《吉林大学》2017年硕士论文　论文类型：学位论文

【摘要】：在现代社会中,身份认证的安全性是人们越来越关注的一个重要方向。而利用人体的生物特征信息来识别身份的技术是目前安全性能最高的一种技术,在生物信息识别技术当中,虹膜识别是一种识别准确率相当高的识别技术,在稳定性,唯一性,可接受性等多个方面比其他生物识别技术具有明显的优势。作者所在的课题小组对虹膜识别已经进行了多年的研究,虹膜识别的过程中有质量评价,虹膜定位,虹膜图像归一化,虹膜图像增强,虹膜特征提取,虹膜特征对比等多个过程,每个过程又有着不同的算法,例如微积分圆模板检测定位方法,非同心圆定位方法,样条拟合算法等定位方法,基于Gabor滤波器的特征提取方法,小波过零点的特征提取方法,SVM分类方法等。我们需要一个虹膜识别平台对现有的算法进行组合,测试,以期在实验室已有算法当中,得到一个识别效率最高的虹膜算法流程。本文所设计并实现的虹膜识别实验平台使各个算法实现模块化开发,将算法封装为一个个保留接口的动态链接库,实验流的整个操作过程,细化为对每个流程节点的算法模块的操作,在平台上可以轻松的组建指定算法流程的实验过程。使实验流的构建更加的便捷化,柔性化。本文首先对虹膜识别相关的背景做了简单介绍,然后分析了虹膜识别过程中的相关流程节点及其具体算法。对实验平台做了详细的需求分析,并建立了相关模型。然后根据虹膜实验流程中流程节点和算法模块之间的关系,建立N叉树的存储结构,方便对其查询及处理以及流程模板及实验流的构建。分析了算法模块的输入输出信息,规范了算法模块的接口,完成了各个算法模块的封装。然后建立了流程模板结构,在满足流程节点之间的逻辑关系的前提下,最大限度扩展流程模板的自由度。设计并实现了实验流引擎,完成了流程模板向实验流的实例化转化,实现了实验流的自动运行。利用建立的实验平台可以实现任何一个单独算法模块运行的结果测试,还可以在满足流程节点逻辑关系的前提下,随意的组合已有的流程节点,并且可以根据用户的需求,自定义添加流程节点,可以实现用户自定义的实验流过程的结果测试。流程节点下方的不同算法模块可以在实验流中相互替换,在平台上可以实现同一流程下,同一流程节点下方不同算法模块的识别率比较。最后,根据平台测试出的实验流算法模块,我们构建了虹膜识别身份认证系统,将实验平台的测试结果向实际应用进行转化。
[Abstract]:In modern society, the security of identity authentication is an important direction that people pay more and more attention to. In the biometric recognition technology, iris recognition is a kind of recognition technology with high recognition accuracy. It is stable and unique. Acceptability and other aspects have obvious advantages over other biometrics. The author's research group has been studying iris recognition for many years, and the iris recognition process has quality evaluation, iris location, and so on. The process of iris image normalization, iris image enhancement, iris feature extraction, iris feature contrast and so on, each process has its own algorithm, such as calculus circle template detection and localization method, non-concentric circle localization method, etc. We need an iris recognition platform to combine and test the existing algorithms, such as spline fitting algorithm, feature extraction method based on Gabor filter, feature extraction method over 00:00 wavelet and SVM classification method, etc. In order to obtain the most efficient iris algorithm flow in the laboratory, the iris recognition experiment platform designed and implemented in this paper makes each algorithm realize modularization development. The algorithm is encapsulated into a dynamic link library with a reserved interface, and the whole operation process of the experimental flow is refined into the operation of the algorithm module of each process node. It is easy to set up the experimental process of specifying the algorithm flow on the platform. It makes the construction of the experimental flow more convenient and flexible. Firstly, this paper briefly introduces the background of iris recognition. Then it analyzes the relevant flow nodes and their specific algorithms in the process of iris recognition, and makes a detailed demand analysis of the experimental platform, and establishes the relevant model. Then, according to the relationship between the flow node and the algorithm module in the iris experimental process, The storage structure of N-tree is established to facilitate the query and processing of N-tree, as well as the construction of flow template and experimental flow. The input and output information of algorithm module is analyzed, and the interface of algorithm module is standardized. The encapsulation of each algorithm module is completed. Then the structure of process template is established, and the degree of freedom of process template is maximized under the premise of satisfying the logical relationship between process nodes. The experimental flow engine is designed and implemented. The instantiation of the flow template to the experimental flow is completed, and the automatic running of the experimental flow is realized. The result test of any single algorithm module can be realized by using the established experimental platform. Under the premise of satisfying the logic relation of the process node, we can combine the existing process nodes at will, and add the process nodes according to the needs of the users. Different algorithm modules under the flow node can be replaced in the experiment flow, and the same flow can be implemented on the platform. Finally, according to the experimental flow algorithm module tested by the platform, we build the iris identification authentication system, and transform the test results of the experimental platform to the actual application.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】