本科毕业设计(论文) 题目名称：
学院：
专业年级：
学生姓名：
班级学号：
指导教师：
二O一O年月日
目录
摘要 ................................................................................................................................ III ABSTRACT ....................................................................................................................... IV 第一章绪论 (1)
1.1 研究背景 (1)
1.2 研究现状综述 (2)
1.2.1 特征点提取研究现状 (2)
1.2.2 特征点匹配研究现状 (4)
1.3 研究内容 (5)
1.4 论文组织结构 (6)
第二章图像特征点提取算法研究 (8)
2.1 Harris特征点提取算法 (8)
2.1.1 Harris特征点检测原理 (8)
2.1.2 Harris特征点检测步骤 (9)
2.1.3 实验分析 (9)
2.2 SIFT特征点提取算法 (12)
2.2.1 SIFT特征点提取算法原理及步骤 (12)
2.2.2 实验分析 (16)
2.3 本章小结 (19)
第三章改进的H/S特征点提取算法 (20)
3.1 H/S算法分析 (20)
3.1.1 H/S算法提出的可行性分析 (20)
3.1.2 H/S算法结合方案研究 (21)
3.2 H/S算法实现 (22)
3.2.1 多尺度特征点检测 (22)
3.2.2 特征点描述 (23)
3.3 实验分析 (25)
3.3.1 算法评价标准 (25)
3.3.2 分情况特征点提取结果分析 (25)
3.4 本章小结 (28)
第四章基于H/S的特征点匹配方法 (29)
4.1 相似三角形方法局限性分析 (29)
4.1.1 三角形匹配原理 (29)
4.1.2 三角形相似方法局限性 (31)
4.2 改进的相似三角形匹配方法实现 (31)
4.2.1 三角形选取方案研究 (32)
4.2.2 相似三角形检索方法改进 (33)
4.2.3 多模板改进方法分析 (35)
4.3 实验结果与分析 (36)
4.4 本章小结 (39)
第五章工作总结与展望 (40)
5.1 工作总结 (40)
5.2 工作展望 (41)
致谢 (42)
参考文献 (43)
摘要
图像匹配是图像处理领域的基础课题，在图像拼接、全景视图、对象识别等计算机视觉方面有广泛应用。图像匹配分为基于灰度的方法和基于特征的方法，前者简单易行，但算法时间复杂度高，难以处理图像存在旋转和尺度变化等情况，后者在适应性和速度方面有很大优势，但图像特征的提取以及如何建立特征之间的对应关系一直是这类方法的研究难题。本文提出了一种新的特征点检测算法，并根据三角形相似原理，提出一种改进特征点匹配方法。
本文首先对两种特征点提取算法(Harris算法和SIFT算法)进行分析比较，Harris 角点检测算法效率高，但对尺度变化和抗噪性的鲁棒性差，SIFT算法对尺度变化及抗噪性鲁棒性好，但算法时间复杂度高，且对图像纹理要求高，本文结合两种算法的优点，提出了一种新的H/S特征点提取算法，该算法特征点检测效率高，并且尺度不变性及抗噪性得到很大改善。
在采用新的H/S特征点提取算法得到图像中特征点后，对特征点匹配方法进行研究。传统的三角形相似匹配方法虽然实现简单，鲁棒性好，但时间复杂度高，针对此局限性，本文提出了一种改进三角形匹配方法，将传统方法被动搜索相似三角形变为在复数空间下选择基线后主动构造相似三角形，并利用特征点组成基线向量对应三角形组得到多个模板在实物图中匹配，提高了改进算法的实时性和稳定性。
关键词特征点，H/S算法，机器视觉，相似三角形检索
ABSTRACT
Image matching is a basic problem in the field of image processing, it has wide applications in computer vision, such as image stitching, panoramic view, object recognition and so on. It includes gray-based method and feature-based method. The former method is simple, but the time complexity of this algorithm is high, especially it’s difficult to deal with the situations of image rotation and scaling. The latter method is much more easily overcoming the difficulties which encountered by the former method, but how to create the corresponding relationships between the images’ features is always a difficult problem. A new feature point extracting method is proposed in this thesis, and a new matching method which is based on similar triangles is proposed.
The thesis firstly analyses and compares two kinds of extracting feature points algorithms( Harris algorithm and SIFT algorithm ), Harris algorithm has high efficiency and poor anti-scaling and anti-noise performance, SIFT algorithm has good robustness about anti-scaling and anti-noise, but the time complexity of this algorithm is high, and it has a great demand on image texture. The thesis combines the advantages of these two algorithms, a new H/S algorithm which is used to extract feature points is proposed, this algorithm has well efficiency, meanwhile, the anti-scaling and anti-noise performance is greatly improved.
While the feature points of images are obtained by using the new H/S algorithm, the matching methods are studied. Traditional similar triangles matching method is simple and has good robustness, but the time complexity of this algorithm is high, because of this limitations, the thesis proposes a improved matching method which is based on similar triangles, making the traditional passively searching similar triangles to select the baseline to actively construct similar triangles. The algorithm uses multi-templates to match in the image, the real time ability and stability of the algorithm is improved.
KEY WORDS feature points, H/S algorithm, computer vision, similar triangles retrieval
