指纹自动识别系统中指纹细节特征点提取算法研究

发布时间：2018-06-20 07:11

本文选题：指纹 + 指印　；参考：《西南政法大学》2010年硕士论文

【摘要】： 指纹是手指末节的乳突花纹,其花纹结构较之于手掌面其他部位最为复杂、典型,具有较高的应用价值。具有各人各指不同、终身稳定不变的特性。指纹因其蕴涵大量的人身个体信息,而具有很高的人身识别价值。百多年来,人们通过对指纹不懈的研究和探索,逐步对指纹的特征体系有了清晰的认识,在此基础上,人们对指纹特征提取的算法进行了大量研究,开发出了比较完善的指纹自动识别系统。在指纹识别技术中,指纹特征提取技术是其中一个非常重要的部分。如果指纹特征得不到准确有效的提取,将直接影响指纹特征匹配的质量。许多学者对指纹特征提取方法进行了探讨,其中,主要包括两大类方法,即依据指纹灰度图像直接提取指纹特征的方法和依据二值化后的指纹图像提取特征的方法。实验表明,基于二值化指纹图像提取特征方法所提取到的特征更全面、质量更高。上述方法均为通过对指纹图像进行局域分析来估计指纹方向图,而后对图像进行滤波增强。当系统中输入的指纹源图像质量较低时,其分辨率会大大降低,这些方法难以准确得到局域的方向图估算值,将无法进行有效的滤波增强。实质上,司法领域中所需识别的指纹图像大多遭到严重污染,分辨率较低,因此,现有的特征提取算法大多难以在该领域发挥较大的作用。针对此情况,为增强指纹识别系统处理低质量指纹的能力,在总结相关研究成果的基础上,本文提出了一套新的指纹图像细节特征提取算法。本文分为三个部分,第一部分着重阐述指纹的基础知识和分类概况;第二部分阐述指纹自动识别系统的概况、指纹自动匹配的基本原理以及指纹自动识别系统的内容和特点;第三部分详细阐述指纹细节特征提取的改进算法,该部分从比较基于灰度指纹图像提取细节特征的算法与基于二值化指纹图像提取细节特征的算法入手,着重论述了基于细化二值化指纹图像细节特征点提取的改进算法,以及如何对用该种算法提取到的细节特征中一些典型的伪特征点的去除;结束语部分总结了本文的研究价值和对细节特征提取算法的展望。该套算法分为七个部分:一是指纹图像预处理算法。此算法包括对指纹图像进行规一化、分割、指纹图像的增强三方面的算法;二是指纹图像的二值化。在指纹图像预处理的基础上对指纹图像进行二值化,目的在于将指纹图像转换为0和1表示的图像。三是指纹图像的细化;四是基于细化图像的细节特征提取算法的改进,该部分主要论述了基于纹线端点和分叉点提取特征的改进算法;五是伪特征点的去除,主要内容阐述了毛刺、短脊的删除算法;六是论述了指纹特征点的影响因素,主要论述了阈值和二值化对所提取的特征点的影响;第七部分主要阐述了该算法的效能。本文所述新的提取算法能够在选定的局部窗口内提取到质量比较高的细节特征,较其他算法有所改进,但对于更大面积的低质量指纹图像区域或者整幅指纹图像,本文所述算法暂不能实现细节特征的有效提取,需要进一步结合指纹自身的特点,不断扩展算法的处理窗口,使之具备从宏观上提取到高质量指纹细节特征的能力。这些将是作者继续努力研究的方向。
[Abstract]:Fingerprint is the mastoid pattern of the end of the finger. Its pattern structure is the most complex and typical of the other parts of the palmar surface. It has a high value of application. It has the characteristics of the stability and stability of each person's fingers. The fingerprint has a high personal personal recognition value because it contains a lot of personal information. The unremitting research and exploration of the fingerprint have a clear understanding of the fingerprint feature system. On this basis, people have done a lot of research on the algorithm of fingerprint feature extraction, and developed a relatively perfect automatic fingerprint recognition system. In fingerprint recognition technology, fingerprint feature extraction is one of the most important parts. Fingerprint features can not be extracted accurately and effectively, which will directly affect the quality of fingerprint matching. Many scholars have discussed fingerprint feature extraction methods, including two major methods, namely, the method of extracting fingerprint features directly according to the gray level of fingerprint and the method of extracting features based on the two valued fingerprint images. Experiments show that the extracted features based on the two valued fingerprint image extraction method are more comprehensive and of higher quality.
The above methods all estimate the fingerprint image by local analysis of the fingerprint image, and then filter and enhance the image. When the quality of the fingerprint source image is low, the resolution will be greatly reduced. These methods are difficult to get the estimation value of the local pattern accurately, and the effective filtering will not be enhanced. Most of the fingerprint images needed to be identified in the field of justice are seriously polluted and the resolution is low. Therefore, most of the existing feature extraction algorithms are difficult to play a major role in this field.
In this case, in order to enhance the ability of fingerprint recognition system to deal with low quality fingerprints, a new algorithm for extracting details of fingerprint images is proposed in this paper. This paper is divided into three parts. The first part focuses on the basic knowledge and classification of fingerprints, and the second part describes the automatic fingerprint recognition. The basic principle of automatic fingerprint matching and the content and characteristics of fingerprint automatic identification system are introduced in the third part, and the improved algorithm for extracting fingerprint details is described in detail. This part starts with the algorithm based on the comparison of the details based on the gray level fingerprint image and the algorithm based on the two value fingerprint image extraction. This paper mainly discusses the improvement algorithm based on thinning two value fingerprint image details feature point extraction, and how to remove some typical pseudo feature points in the detail features extracted with this algorithm. The end language summarizes the research value of this paper and the prospect of the Extraction Calculation of detail features. The algorithm is divided into seven parts: 1 It is a fingerprint image preprocessing algorithm. This algorithm includes three aspects of the fingerprint image, the two is the two value of the fingerprint image. The fingerprint image is two valued on the basis of the fingerprint image preprocessing. The purpose is to convert the fingerprint image to the 0 and 1 image. Three is the fingerprint image. The four is the improvement of the detail feature extraction algorithm based on the thinning image. This part mainly discusses the improvement algorithm based on the feature extraction based on the line end point and the bifurcation point; five is the removal of the pseudo feature points, the main content describes the burr, the deletion algorithm of the short ridge, and six discusses the influence factors of the fingerprint feature points, and mainly discusses the threshold value. The two part is the influence on the extracted feature points; the seventh part mainly expounds the effectiveness of the algorithm.
The new extraction algorithm described in this paper can extract high quality details in selected local windows, and is better than other algorithms. However, for the larger area of the low quality fingerprint image region or the whole fingerprint image, the algorithm described in this paper can not achieve the effective extraction of the details for the time being. It needs to be further combined with the fingerprint self. The characteristics of the body, and continue to expand the processing window of the algorithm, so that it has the ability to extract the detailed features of high quality fingerprints from the macro level. These will be the direction of the author's continued efforts.
【学位授予单位】：西南政法大学
【学位级别】：硕士
【学位授予年份】：2010
【分类号】：D918.91

【相似文献】