首页 | 本学科首页   官方微博 | 高级检索  
     

基于POS数据约束的ADS40影像SIFT匹配方法
引用本文:孟俊俊,马慧云,石宁,张云生,邹峥嵘. 基于POS数据约束的ADS40影像SIFT匹配方法[J]. 信息工程大学学报, 2014, 15(3): 370-374
作者姓名:孟俊俊  马慧云  石宁  张云生  邹峥嵘
作者单位:中南大学地球科学与信息物理学院,湖南长沙410083
基金项目:国家自然科学基金资助项目(41201472);国家863计划资助项目(2012AAl20801);中国博士后科学基金资助项目(2012M511763);中央高校基本科研业务费专项资金资助项目(2012QNZT078).
摘    要:连接点提取是ADS40影像数据自动空三的关键技术之一。为加快匹配速度并为高精度子像素匹配提供稳健而相对精确的初值,提出一种基于POS数据约束的ADS40影像SIFT(scale invariant feature transform)特征匹配方法。选取典型ADS40影像数据进行试验,并与传统灰度相关匹配方法及经典的SIFT算法相比,匹配正确率分别提高67.7%和11.99%。试验结果表明,文章提出的方法可提高ADS40影像匹配的成功率。由于匹配点能够均匀分布,因此文章方法能够为ADS40影像自动空三工作提供有效连接点。

关 键 词:ADS40  影像匹配  SIFT  近似核线

ADS40 Image Matching Method Based on SIFT Features Constrained by POS Data
MENG Jun-jun,MA Hui-yun,SHI Ning,ZHANG Yun-sheng,ZOU Zheng-rong. ADS40 Image Matching Method Based on SIFT Features Constrained by POS Data[J]. , 2014, 15(3): 370-374
Authors:MENG Jun-jun  MA Hui-yun  SHI Ning  ZHANG Yun-sheng  ZOU Zheng-rong
Affiliation:(Department of Surveying Engineering & Geo-Informatics, Central South University, Changsha 410083, China)
Abstract:Tie point based on image matching is one of the key technologies of automatic aerial trian- gulation in processing ADS40 image data. To speed up image matching and provide a robust and relatively accurate initial value for high-precision sub-pixel matching, a SIFT feature match method constrained by POS data is proposed. A pair of typical ADS40 images is used to perform the proposed method. Compared with traditional cross-correlation and SIFT feature matching method, the number of matches increases 8.3 and 0.4 times respectively, while the accuracy rate improves by 67.7% and 11.99%. The results reveal that the matching success rate for ADS40 image could be increased by the proposed method. Because of the even distribution of matches, the proposed method is believed to be possible to provide valid tie points for automatic aerial triangulation of ADS40 images.
Keywords:ADS40  image matching  SIFT  qusi-epipolar line
本文献已被 维普 等数据库收录!
点击此处可从《信息工程大学学报》浏览原始摘要信息
点击此处可从《信息工程大学学报》下载全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号