基于DSP的手背静脉识别系统研究

发布时间：2018-03-01 04:46

本文关键词： 手背静脉识别 Hausdorff距离不变矩特征融合 DSP　出处：《吉林大学》2016年硕士论文　论文类型：学位论文

【摘要】：随着社会不断地发展,人们对安全的可靠性要求越来越高,许多场合都需要对人的身份进行认证与识别,因此生物特征识别技术显得越来越重要。而人体手背静脉作为一种崭新的生物特征识别技术,与视网膜技术、指纹技术、虹膜技术等相比,它具有内部特征、活体识别、安全等级高和非接触式等优点,表现出明显优势,并且手背静脉识别技术处于刚刚起步的阶段,所以人们对手背静脉识别技术的深入研究非常有必要和实用意义。目前,手背静脉识别一般都是利用PC机来处理,但这样的识别系统体积过于庞大,而利用DSP来进行手背静脉识别就能较好的解决此类问题。本文使用的DSP芯片型号为TI公司推出的TMS320DM642,该数字图像处理芯片是专用于图像以及视频处理方面的,并且主频最高可以达到300Mhz。国内手背静脉识别产品的核心技术大多数都是从韩国NEXTERN公司购买的,并且将DSP应用于生物特征识别设备中,可以有效降低成本,便于产品化,所以本文设计了一套基于DSP的手背静脉识别系统。本文采用Mohamed Shahin博士手背静脉图像库,共选取200幅手背静脉图像,对基于DSP的手背静脉识别系统进行深入研究,本文主要的研究内容为:(1)针对现有手背静脉识别利用的手背静脉特征较少,识别率较低等缺点,本文提出了一种基于特征点距离的手背静脉特征融合的方法。该方法首先提取7个不变矩,再提取手背静脉的端点及交叉点,将几何特征和局部特征相融合,实验结果准确率可达96.75%。(2)本文介绍了手背静脉识别系统的硬件平台,本系统的硬件平台包括系统的选型、系统的结构及各模块设计。本文通过对比几款DSP,最终选取DM642芯片作为系统的核心处理器。(3)本文设计了手背静脉识别系统的软件框架,并且介绍了软件的开发环境、DSP/BIOS操作系统等。为本文成功完成基于DSP的手背静脉识别算法移植的后续工作做出铺垫。(4)本文设计了一套基于DSP的手背静脉识别系统,将本文研究的两种手背静脉识别算法移植到DSP上,该算法能成功判断是否为同一个人的手背静脉,实验结果准确率可达93.5%,验证了本文算法的实用性和可靠性。本文利用Hu的7个不变矩方法与豪斯多夫距离法进行算法融合,成功完成了基于DSP的手背静脉识别算法移植。
[Abstract]:With the continuous development of society, people demand more and more security reliability, many occasions need to authenticate and identify the identity of people. Therefore, biometric identification technology is becoming more and more important. As a new biometric recognition technology, human dorsal hand vein has internal features and in vivo recognition compared with retina technology, fingerprint technology, iris technology, etc. The advantages of high safety grade and non-contact type show obvious advantages, and the technology of dorsal hand vein recognition is just in its infancy, so it is very necessary and practical to study deeply the technology of dorsal hand vein recognition. The back of the hand vein recognition is usually processed on a PC, but this recognition system is too large. In this paper, the model of DSP chip is TMS320DM642produced by TI, and the digital image processing chip is specially used in image and video processing. Most of the core technologies of domestic hand back vein recognition products are purchased from NEXTERN Company of South Korea, and the application of DSP in biometric equipment can effectively reduce the cost and facilitate the production. In this paper, we design a set of dorsal hand vein recognition system based on DSP. In this paper, we select 200 back hand vein images by using Mohamed Shahin image library, and deeply study the DSP based dorsal hand vein recognition system. The main research content of this paper is: (1) aiming at the shortcomings of the existing dorsal hand vein recognition, such as fewer features of the dorsal hand vein, lower recognition rate and so on. In this paper, a method of feature fusion of dorsal hand vein based on the distance of feature points is proposed. Firstly, seven invariant moments are extracted, then the endpoints and intersections of the dorsal hand vein are extracted, and the geometric features and local features are fused. This paper introduces the hardware platform of the hand back vein recognition system. The hardware platform of the system includes the selection of the system. The structure of the system and the design of each module. By comparing several DSP, the DM642 chip is selected as the core processor of the system.) the software framework of the hand back vein recognition system is designed in this paper. This paper also introduces the software development environment, such as DSP / BIOS operating system. This paper designs a DSP based hand back vein recognition system for the successful completion of the follow-up work of hand back vein recognition algorithm transplantation based on DSP. By transplanting two algorithms of dorsal hand vein recognition to DSP, the algorithm can successfully judge whether the vein is the same person's dorsal hand vein. The accuracy of the experiment can reach 93.5a, which verifies the practicability and reliability of this algorithm. This paper uses Hu's seven moment invariant method and Hausdorf distance method to fuse the algorithm, and successfully completes the transplantation of the hand back vein recognition algorithm based on DSP.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41

【参考文献】