首页 > 论文 > 光学学报 > 33卷 > s2期(pp:s211001--1)

基于多尺度特征的遥感图像密集匹配方法

Method of Remote Sensing Images Dense Matching Based on Multi-Scale Features

  • 摘要
  • 论文信息
  • 参考文献
  • 被引情况
  • PDF全文
分享:

摘要

提出一种利用多尺度特征的无人飞艇遥感平台获取的序列航拍图像生成密集匹配视差图像的方法。运用尺度不变特征变换(SIFT)算法从两幅相邻图像中提取特征点,基于欧氏距离进行粗匹配,并通过缩小搜索范围提高粗匹配效率,再通过随机抽样一致性(RANSAC)算法估计基础矩阵,利用对极几何约束关系剔除误匹配,进行精确匹配,提高匹配的稳健性和精度,然后利用区域生长算法进行特征点密集匹配生成相应视差图像。实验表明算法在保持稳健性的同时,可以降低时间复杂度,获得相当规模的密集匹配点,得到良好的可视效果。

Abstract

The algorithm adopting multi-scale features to generate parallax images with dense matching for image sequences collected from unmanned airship platform is proposed. Scale invariant feature transform (SIFT) algorithm is adopted to extract features from adjacent images firstly. The initial matches based on Euclidean distance are carried out, and the efficiency of feature matching is improved by reducing the search area. The random sample consensus (RANSAC) algorithm is adopted to estimate the fundamental matrix, and the epipolar geometry constrain is used to delete the wrong matches to improve the robustness and accuracy of feature matching. The region growing algorithm is adopted to carry out dense matching and produce parallax images. Experimental results indicate that the proposed algorithm keeps robustness, reduces the time complexity, generates lots of dense feature matches and gets good visual effects.

广告组1 - 空间光调制器+DMD
补充资料

中图分类号:TP274.2

DOI:10.3788/aos201333.s211001

所属栏目:成像系统

基金项目:国家科技支撑计划(2012BAH31B01)、国家自然科学基金(40601081, 41071255)

收稿日期:2013-05-21

修改稿日期:2013-06-28

网络出版日期:--

作者单位    点击查看

胡少兴:北京航空航天大学机械工程及自动化学院, 北京 100191
王惟达:北京航空航天大学机械工程及自动化学院, 北京 100191
柴进:北京航空航天大学机械工程及自动化学院, 北京 100191
张爱武:首都师范大学三维信息获取与应用教育部重点实验室, 北京 100037

联系人作者:胡少兴(husx98@163.com)

备注:胡少兴(1972—),男,博士,副教授,主要从事计算机视觉、智能控制及光电产品开发与设计等方面的研究。

【1】Chen Le, Lü Wenge, Ding Shaohua. A survey of corner detection algorithms [J]. Techniques of Automation and Applications, 2005, 24(5): 1-4.
陈乐, 吕文阁, 丁少华. 角点检测技术研究进展[J]. 自动化技术与应用, 2005, 24(5): 1-4.

【2】D G Lowe. Distinctive image features from scale-invariant keypoints [J]. International J Computer Vision, 2004, 60(2): 91-110.

【3】J Koenderink. The structure of images [J]. Biological Cybernetics, 1984, 50(5): 363-396.

【4】T Lindeberg. Scale-space for discrete signals [J]. IEEE Trans Pattern Analysis and Machine Intelligence, 1990, 12(3): 234-254.

【5】J Babaud, A P Witki, M Baudin, et al.. Uniqueness of the Gaussian kernel for scale-space filtering [J]. IEEE Trans Pattern Analysis and Machine Intelligence, 1986, 8(1): 26-33.

【6】K Mikolahczyk. Detection of Local Features Invariant to Affine Transformations [D]. Grenoble: Institute National Polytechnique de Grenoble, 2002. 37-40.

【7】D G Lowe. Object recongnition from local scale-invariant features [C]. Proceedings of the 7th International Conference on Computer Vision, 1999, 2: 1150-1157.

【8】Wang Xiaohua, Fu Weiping, Liang Yuanyue. A method for improving SIFT feature matching efficiency [J]. Mechanical Science and Technology for Aerospace Engineering, 2009, 28(9): 1252-1255.
王晓华, 傅卫平, 梁元月. 提高SIFT特征匹配效率的方法研究[J]. 机械科学与技术, 2009, 28(9): 1252-1255.

【9】Liu Xiaojun, Yang Jie, Sun Jianwei, et al.. Image registration approach based on SIFT [J]. Infrared and Laser Engineering, 2008, 37(1): 156-160.
刘小军, 杨杰, 孙坚伟, 等. 基于改进SIFT的图像配准算法[J]. 红外与激光工程, 2008, 37(1): 156-160.

引用该论文

Hu Shaoxing,Wang Weida,Chai Jin,Zhang Aiwu. Method of Remote Sensing Images Dense Matching Based on Multi-Scale Features[J]. Acta Optica Sinica, 2013, 33(s2): s211001

胡少兴,王惟达,柴进,张爱武. 基于多尺度特征的遥感图像密集匹配方法[J]. 光学学报, 2013, 33(s2): s211001

您的浏览器不支持PDF插件,请使用最新的(Chrome/Fire Fox等)浏览器.或者您还可以点击此处下载该论文PDF