《基于数字图像处理的sift算法的研究》由会员分享，可在线阅读，更多相关《基于数字图像处理的sift算法的研究（35页珍藏版）》请在金锄头文库上搜索。
1、太原科技大学毕业设计（论文）太原科技大学毕 业 设 计（论 文）设计(论文)题目：基于数字图像处理的 SIFT 算法的研究姓 名 学院（系） 电子信息工程学院专 业 测控技术与仪器 年 级 指导教师 2016 年 6 月 日太原科技大学毕业设计（论文）太原科技大学毕业设计（论文）I基于数字图像处理的 SIFT 算法的研究电子信摘要：人类获取信息的方式都包括触觉、嗅觉、视觉、听觉等，而视觉更是人类获取外界信息最重要的方式,受视觉的启发,人类开始着手研究如何让机器拥有一套类似于视觉的获取外界信息的渠道,于是在二十世纪五十年代计算机视觉领域应运而生,无数人都在这一门崭新的领域内倾尽心血进行探索,随后
2、,Marr 教授在总结前人的基础上提出了最为具体的计算机视觉理论框架体系。在计算机视觉理论当中,关于图像特征点的提取就是一个经典难题,为了更好的提取稳定的特征点,学者们做出了众多的探索,SUSAN 算子、Harris算子和 SIFT 算子等特征点的提取方法就是在这不断探索中诞生的一小部分。在众多特征点的提取方法中,SIFT 算法因其先天具有的多样优势从众多方法之中脱颖而出并且得到了较为广泛的应用,但是 SIFT 算法并没有做到尽善尽美，在实际运用中还存在着各种缺陷,计算维度过大导致运用 SIFT 算法在处理特征点的提取与匹配时无法满足我们对于算法实时性的要求,之后不断有学者对 SIFT 算法的
3、缺点提出改进来加快原始 SIFT 算法的运算速度,常见的 PCA-SIFT、SURF 等改进算法就是这一时期基于 SIFT 算法提出改进所获得的果实。本文首先简单介绍了计算机视觉和 SIFT 算法的发展背景和现状,在特征点的提取之中将 SIFT 算子和 SUSAN、 Harris 算子进行横向比较突出 SIFT 算法的优越性,然后详细介绍SIFT 算子的构建过程并通过仿真来展示特征点的提取效果,针对 SIFT 算法存在的实时性不强的缺点,本文将通过纵向比较的方法来简要介绍近年来学者们对于 SIFT 算法改进中运用较为广泛的的 PCA-SIFT 算法和 SURF 算法,随后通过使用计算机来模拟仿
4、真实验展示两种改进算法对于特征点的提取效果，并具体说明在算法的改进之中所取得的成果。关键词: SIFT， 特征点， 计算机视觉The SIFT algorithm based on digital image processing researchCollege of Electronical and Information Engineering measurement and control YanTengfei Director：WangQinggangAbstract：Human access to information in a way that includes a touch,
5、 smell, sight, 太原科技大学毕业设计（论文）IIhearing and, while vision is for people to obtain information from outside the most important way, 太原科技大学毕业设计（论文）IIIinspired by the vision, humans began to begin to study how to make the machines have a similar to visual access to outside information channels, so in th
6、e field of computer vision, in the fifties of the 20th century came into being, millions of people are in the a new field of introverted try hard to explore. Subsequently, Professor Marr on the basis of summarizing the proposed the specific computer vision theory framework. In the theory of computer
7、 vision, on the image feature points extraction is a classical problem, in order to better extract stable feature points, scholars have made many exploration, SUSAN operator and Harris operator and SIFT operator feature point extraction method is in this continuously penetrating the cable was born i
8、n a small part. In many feature points extraction methods, SIFT algorithm because of its innate has various advantages stand out from many methods and get more extensive application, but the SIFT algorithm did not achieve perfection. Various defects still exist in the practical application, dimensio
9、n calculation through to use the SIFT algorithm in processing of feature points extraction and matching cannot meet we for real-time requirements. Later, scholars of SIFT algorithm improve shortcoming to accelerate the speed of operation of the original SIFT algorithm, common PCA-SIFT, SURF and impr
10、oved algorithm is the period based on SIFT algorithm is proposed to improve the fruit.This paper introduces the development background and present situation of computer vision and SIFT algorithm, the superiority of the more prominent SIFT algorithm in transverse feature extraction in SUSAN, SIFT ope
11、rator and Harris operator, and then introduces the construction process of the SIFT operator and the simulation to demonstrate the extraction of feature points, according to the real-time SIFT algorithm the existence of weak, this will be through the longitudinal comparison method to introduce schol
12、ars in recent years for the more extensive PCA-SIFT algorithm and SURF algorithm with the improved SIFT algorithm, and then through the use of computer simulation experiments show the two algorithms for the extraction of feature points, and illustrates the fruit in the algorithm.Key Words： SIFT, feature point, computer vision太原科技大学毕业设计（论文）IV目录第 1 章 绪论 .11.1 研究背景及意义 . 11.2 国内外研究现状 . 2 1.3 本文结构 . 2 第 2 章 SIFT 算法 .3 2.1 常见特征点提取算法的比较： .3 2.2 SIFT 算子的建立过程 . 5 2.2.1 构建尺度空间 .5 2.2.2 特征点的选取 . 10 2.2.3 特征点方向的确定 .132.2.4 生成特征点描述子 . 14 2.3 特征点的匹配 .
