万有引力场方法在人耳图像识别中的应用研究

发布时间：2018-04-27 05:31

本文选题：人耳识别 + 万有引力场　；参考：《重庆大学》2007年硕士论文

【摘要】： 人耳识别是生物特征识别技术研究中的一个新兴领域,它旨在根据人的外耳特征进行身份识别。人耳识别可以作为其他生物识别技术的有益补充,也可以单独应用于一些个体识别场合。目前人耳识别技术在国内外尚处于初步探索研究阶段,还没有形成较为完善的理论体系。作为一种生物特征识别技术,人耳识别需要借鉴其他的生物识别技术和处理方法,同时人耳具有其特殊的生理结构和生理位置,必然有其独特的处理方法和技术与之相适应。本文以人耳这一特殊的生物特征为研究对象,并在前人所做工作的基础上,以人耳图像万有引力场转换、特征提取等关键环节为研究重点,力图改善和丰富人耳识别这种新兴的生物特征识别技术,以期对后续研究人员的工作进展有所启发。众所周知,确定特征空间的最终目标是在降低原有的维度空间的同时,又能保持正确分类识别的能力。为了达到这个目标,在人耳生物识别研究领域中,一种新的万有引力场转换理论把图像看作一系列相互吸引的像素粒子,这些粒子就作为万有引力场源。伴随力场产生的还有一个标量势能场,在势能场中,人耳轮廓具有光滑的表面,就像很多由山脊连接的山峰一样。这里的山峰对应的就是势能阱,类似地山脊就对应着势能通道。本文着重描述了如何利用万有引力场转换算法来提取力场特征,利用场线方向来定位势能通道和势能阱,这些势能通道和势能阱最终将构成人耳的特征信息。为了进一步说明力场转换具有可逆性、良好的亮度灵敏性及定位迅速等优点,本文利用自己建立的人耳图像库完成了实验。并通过对比硬C-均值,模糊C-均值和改进模糊C-均值聚类算法的实验结果,来说明力场转换理论应用于人耳识别的有效性。本文不仅证明了人耳识别是一种有着极大发展潜力的生物识别技术,还证明了力场转换具有很好的鲁棒性,特别是对噪声的容忍度。另外人耳轮廓不必从背景中提取出来,这点也是它的优势所在。总的来说,本文主要从力场转换理论的几个关键环节进行了有益的探索与研究,得到了较好的实验结果,为人耳识别技术的发展和应用提供了有价值的研究思路。
[Abstract]:Ear recognition is a new research field in biometric recognition. Ear recognition can be used as a useful supplement to other biometric techniques and can also be used in individual recognition situations. At present, the human ear recognition technology is still in the initial stage of exploration and research at home and abroad, and has not yet formed a relatively complete theoretical system. As a biometric recognition technology, human ear recognition needs to draw lessons from other biometric identification techniques and processing methods. At the same time, human ear has its special physiological structure and position, so it must have its unique processing methods and techniques to adapt to it. In this paper, the special biological characteristics of human ear are taken as the research object, and on the basis of the previous work, the key links such as the transformation of gravitational field and feature extraction of human ear image are taken as the focus of the research. This paper tries to improve and enrich the new biometric recognition technology of human ear recognition, so as to enlighten the future work of researchers. It is well known that the ultimate goal of determining feature space is to reduce the original dimension space while maintaining the ability of correct classification and recognition. In order to achieve this goal, in the field of human ear biometrics, a new universal gravitational field transformation theory regards the image as a series of pixel particles that attract each other, and these particles are regarded as the source of gravitational field. The force field is accompanied by a scalar potential energy field in which the human ear has a smooth surface like many peaks connected by ridges. The peak here corresponds to a potential energy trap, similar to a ridge corresponding to a potential energy channel. This paper mainly describes how to extract the force field features by using the universal gravitational field transformation algorithm, and how to locate the potential energy channels and potential energy wells by using the field line direction. These potential energy channels and potential energy wells will eventually form the characteristic information of the human ear. In order to further illustrate the advantages of force field conversion, such as reversibility, good brightness sensitivity and rapid localization, this paper has completed the experiment by using the ear image database established by ourselves. By comparing the experimental results of hard C- mean, fuzzy C- mean and improved fuzzy C- mean clustering algorithm, the validity of applying force field transformation theory to ear recognition is demonstrated. This paper not only proves that ear recognition is a biometric technology with great potential, but also proves that the force field conversion has good robustness, especially the tolerance to noise. In addition, the contour of the ear does not have to be extracted from the background, which is also its advantage. In general, this paper has carried on the beneficial exploration and the research mainly from several key links of the force field conversion theory, has obtained the better experimental result, has provided the valuable research thought for the human ear recognition technology development and the application.
【学位授予单位】：重庆大学
【学位级别】：硕士
【学位授予年份】：2007
【分类号】：D918.2

【相似文献】