基于图像质量评价的掌纹识别算法研究

发布时间：2018-01-07 17:12

本文关键词：基于图像质量评价的掌纹识别算法研究　出处：《北京交通大学》2016年硕士论文　论文类型：学位论文

【摘要】：掌纹识别技术因其包含信息量大、采集设备简单、用户易于接受等优点,近年来得到了广泛的研究,并成功应用于公安系统、医学研究、社会福利等方面。随着非接触式掌纹采集系统的发展,图像的质量成为一个不可忽略的问题。易用性的提高必将降低图像的质量。而在众多影响图像质量的因素中,离焦模糊是一种极为重要的图像失真。而传统的模糊掌纹识别算法主要通过提取掌纹的稳定性特征进行识别,识别率较低。本文在引入掌纹图像质量评价的方法,并在此基础上提出了LMCP (局部多层对比模式)和log-Gabor数据融合的方法,主要工作如下：(1)引入掌纹图像质量评价。该方法基于Daugman算法,首先对掌纹图像进行拉普拉斯锐化,然后计算图像的质量分数,通过设定的阈值对清晰图像和模糊图像进行区分,去除严重模糊的掌纹图像。(2)提出LMCP和log-Gabor数据融合的掌纹识别算法。首先对LMCP算法和log-Gabor算法进行了分析,实验证明了两种方法的有效性。然后对LMCP和log-Gabor算法进行数据融合,在提取LMCP特征和log-Gabor特征的基础上,进行分数级融合,再通过K-近邻分类器完成匹配识别。仿真结果证明了数据融合的方法使得识别率得到了提高。
[Abstract]:Palmprint recognition technology has been widely studied in recent years because of its advantages such as large amount of information, simple acquisition equipment, easy to accept by users, and has been successfully applied to public security system and medical research. Social welfare. With the development of contactless palmprint acquisition system. Image quality has become a problem that can not be ignored. The improvement of ease of use will reduce the image quality. Defocus blur is a very important image distortion, and the traditional fuzzy palmprint recognition algorithm is mainly by extracting the stability features of palmprint recognition. The recognition rate is low. This paper introduces the method of palmprint image quality evaluation and puts forward the method of LMCP (local multilayer contrast mode) and log-Gabor data fusion. The main work is as follows: 1) the palmprint image quality evaluation is introduced. This method is based on Daugman algorithm. Firstly, the palmprint image is sharpened by Laplace, and then the image quality fraction is calculated. The clear image and the blurred image are distinguished by the set threshold. The palmprint recognition algorithm based on LMCP and log-Gabor data fusion is proposed. Firstly, the LMCP algorithm and log-Gabor algorithm are analyzed. Experiments show the effectiveness of the two methods. Then the LMCP and log-Gabor algorithms are fused to extract LMCP and log-Gabor features. The fractional-level fusion is carried out, and then the matching recognition is accomplished by the K-nearest neighbor classifier. The simulation results show that the recognition rate is improved by the method of data fusion.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41

【相似文献】