基于非高斯Gabor滤波器的掌纹识别算法研究

发布时间：2018-01-29 13:23

本文关键词： 掌纹识别特征提取 LBP 相位特征 Gabor滤波器非高斯Gabor滤波器　出处：《昆明理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：近年来,随着生物识别技术的广泛发展,相关的研究人员对此表现出极大的关注度,是近年来研究的热点之一。生物识别技术有其自身的特点,稳定性高、安全并且方便,从而得到了快速的发展。掌纹是用来描述手掌内部表面上所有纹线的统称,它包括主线、皱褶和乳突纹等。与其他生物特征识别方法相比,掌纹特征信息更加丰富、图像采集简便、使用的用途广泛,受众接受度高,能够有效的识别,所以掌纹识别用来作为一种身份识别方法,得到了普遍性的认可。掌纹识别经过十多年的研究,在理论和实践方面都有了一定的基础。在掌纹图像的预处理方面趋于成熟。由于掌纹表面纹线深浅不一、方向呈现无规律变化,并且导致掌纹图像变化及变形的因数较多,对最终的识别率造成了严重的干扰。本文研究当下一些主流的掌纹识别算法,从掌纹的特征纹理分析、特征融合方向提出精确度高、识别效果好的识别算法,本文的研究如下:1、对文中涉及的掌纹识别预处理算法进行研究分析,提取到掌纹图像的感兴趣区域(ROI);2、提出一种基于非高斯Gabor·滤波器的掌纹识别算法。我们构建出了非高斯Gabor滤波器的函数,对经过预处理的掌纹图像做非高斯Gabor滤波处理,再计算滤波后图像的相位信息,得到相位矩阵。将此相位矩阵进行分块处理,经过相位编码规则计算其相位特征,同时提取到各分块的LBP特征,最后连接各分块的特征向量组成了原始图像的特征向量,最后把该向量送入分类器中进行分类识别;3、在Polyu_Pamprint_Database掌纹库上进行仿真实验,并跟与Gabor滤波器的掌纹算法对比实验,从实验结果上,本文提出的算法更具有高效性,有良好的识别精度。
[Abstract]:In recent years, with the extensive development of biometrics technology, the researchers concerned have shown great attention to it, which is one of the hot research topics in recent years. Biometric recognition technology has its own characteristics and high stability. Palmprint is a general term used to describe all the lines on the inner surface of the palm. It includes the main line, wrinkle and mastoid stripe, etc. Compared with other biometric methods, palmprint is used to describe all the lines on the inner surface of the palm. Palmprint feature information is more abundant, image collection is simple, the use of a wide range of applications, high audience acceptance, can be effectively recognized, so palmprint recognition is used as an identification method. After more than ten years of research, palmprint recognition has a certain theoretical and practical basis. In the palmprint image preprocessing tends to mature. Due to the depth of the palmprint surface lines are different. The direction of the palmprint image changes irregularly, and leads to more factors of palmprint image change and deformation, resulting in serious interference to the final recognition rate. This paper studies some current mainstream palmprint recognition algorithms. From the palmprint feature texture analysis, feature fusion direction proposed a high accuracy, recognition effect of the recognition algorithm, this paper research as follows: 1, the palmprint recognition pre-processing algorithm involved in this paper research and analysis. The region of interest is extracted from palmprint image. 2. A palmprint recognition algorithm based on non-#china_person0# Gabor 路filter is proposed. The function of non-#china_person1# Gabor filter is constructed. The pre-processed palmprint image is processed by non-#china_person0# Gabor filter, then the phase information of the filtered image is calculated, and the phase matrix is obtained. The phase matrix is processed in blocks. The phase features are calculated by the phase coding rules, and the LBP features of each block are extracted at the same time. Finally, the feature vectors connected to each block constitute the feature vectors of the original image. Finally, the vector is sent into the classifier for classification and recognition. 3. The simulation experiment is carried out on Polyu_Pamprint_Database palmprint database and compared with the palmprint algorithm of Gabor filter. The algorithm proposed in this paper is more efficient and has good recognition accuracy.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】