融合指纹与指静脉图像采集系统

发布时间：2018-07-31 18:48

【摘要】：随着信息化技术的快速发展,信息安全问题得到人们的高度重视,传统的身份认证方式具有不易携带、容易被伪造、安全性低等问题而逐渐被新型的生物特征识别技术所取代。相对于目前比较成熟的指纹识别技术,手指静脉识别技术作为新一代的生物特征识别技术,拥有活体检测、内部特性、非接触性等优点,具有更大的发展潜力。在基于指纹或手指静脉的生物特征识别系统中,采集图像的质量直接影响着识别系统的整体性能。因此本文在融合指纹与指静脉图像采集及图像质量增强方面做了如下研究:1.本文首先介绍并分析了目前主流的几种生物特征识别技术及其优缺点,详细介绍了手指静脉识别技术中的图像采集、图像质量评估、图像质量增强以及市场应用方面的相关发展及研究现状。2.本文设计并实现了一种指纹与指静脉图像采集系统,该系统由硬件系统和软件系统组成:(1)硬件系统分为三个模块:基于STM32F407ZGT6芯片的核心控制模块,包括电源接口电路、串口电路、调试接口电路、实时时钟接口电路等组成;一种光源可调、图像可控的手指静脉图像传感器装置,该装置主要包括其外部结构、近红外光源、OV2640图像传感器等;基于FPC1011F3传感器的手指指纹图像采集装置。本文对硬件系统的设计过程和实现原理做出了详细的说明,包括处理器及外围芯片的选型,图像采集装置的整体结构及设计要求,传感器的工作原理及驱动程序设计等。该硬件系统具有成本低、扩展性强、实时性高、采集图像清晰等优点。(2)软件系统实现了两种功能:一种是指纹与指静脉图像采集功能,通过使用USB连接线与硬件系统进行通信,实现对指纹和指静脉图像的采集与文件操作等功能;另一种是静脉图像质量增强功能,利用软件对手指静脉图像进行区域提取、归一化、滤波及增强处理,提高了原始静脉图像的质量。测试结果表明该软件系统具有界面简洁、可控制性高、图像采集速度快、图像增强效果明显等优点。3.为了保证手指静脉采集系统的图像质量,本文提出了一种新的基于混合核函数SVM的手指静脉图像质量分类方法。该算法首先计算手指静脉图像空间域中的一维和二维特征值,将其作为刻画图像质量的特征向量;然后将增强后的图像质量进行质量分类,用于减少对原始图像的主观判断误差;最后利用泛化性能较好的混合核函数SVM方法对大量静脉图像样本进行训练。利用实验验证了该算法的可行性及有效性,说明了混合核函数SVM分类器对手指静脉图像的质量分类具有更好的学习和泛化能力。
[Abstract]:With the rapid development of information technology, people attach great importance to the problem of information security. Traditional authentication methods are difficult to carry, easy to be forged, low security and gradually replaced by the new biometric identification technology. Compared with the current mature fingerprint identification technology, finger vein recognition technology, as a new generation of biometric identification technology, has the advantages of in vivo detection, internal characteristics, non-contact, and has a greater potential for development. In the biometric identification system based on fingerprint or finger vein, the quality of the collected image directly affects the overall performance of the recognition system. In this paper, the fusion of fingerprint and digital vein image acquisition and image quality enhancement do the following research: 1. This paper first introduces and analyzes several popular biometric recognition techniques and their merits and demerits, and introduces in detail the image acquisition and image quality assessment in the finger vein recognition technology. Image quality enhancement and market application related development and research status. In this paper, a fingerprint and finger vein image acquisition system is designed and implemented. The system is composed of hardware system and software system. (1) the hardware system is divided into three modules: the core control module based on STM32F407ZGT6 chip, including power interface circuit, serial circuit, etc. Debugging interface circuit, real-time clock interface circuit and so on, a finger vein image sensor device with adjustable light source and controllable image, which mainly includes its external structure, near infrared light source and OV2640 image sensor, etc. Finger fingerprint image acquisition device based on FPC1011F3 sensor. The design process and implementation principle of the hardware system are described in detail, including the selection of processor and peripheral chips, the overall structure and design requirements of the image acquisition device, the working principle of the sensor and the design of the driver, etc. The hardware system has the advantages of low cost, strong expansibility, high real-time performance, clear image acquisition and so on. (2) the software system realizes two functions: one is fingerprint and finger vein image acquisition function, and the other is to communicate with the hardware system by using USB connection line. The other is the enhancement function of vein image quality, which uses software to extract, normalize, filter and enhance the finger vein image. The quality of the original vein image is improved. The test results show that the software system has the advantages of simple interface, high controllability, fast image acquisition speed and obvious image enhancement effect. In order to ensure the image quality of finger vein acquisition system, a new classification method of finger vein image quality based on mixed kernel function SVM is proposed in this paper. The algorithm firstly calculates one and two dimensional eigenvalues in the spatial domain of the finger vein image as a feature vector to describe the image quality, and then classifies the enhanced image quality. It is used to reduce the subjective judgment error of the original image. Finally, a large number of venous image samples are trained by the hybrid kernel function SVM method, which has better generalization performance. The feasibility and effectiveness of the algorithm are verified by experiments. It is shown that the hybrid kernel function SVM classifier has better learning and generalization ability for the quality classification of finger vein images.
【学位授予单位】：杭州电子科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【相似文献】