摘要
针对多源遥感图像纹理、灰度差异大、数据量大的特点,以及传统配准方法易出现误匹配和低效率问题,提出一种初-精结合的多源遥感图像自动配准方法。首先用最大极值区域检测图像的有效特征区域,再依据区域灰度均方差确定每块区域特征点数量,采用Harris提取区域特征角点。MSER-Harris方法保证了角点分布的均匀和非冗余性。其次对图像作Contourlet变换,在分解的低频子带上构造多尺度高斯组合矩对图像进行初匹配,在高频子带上构造表征图像纹理特征的多方向灰度共生矩完成精匹配,实现同名特征点的配准。对多源遥感图像进行配准实验的结果表明,该方法在特征点数量、分布均匀度及配准精度等方面具有显著的优势,可为后期研究提供参考。
Aiming at addressing the mismatching and low efficiency based on traditional method in multi-source remote sensing image registration due to amount of points,a novel initial-fine automatic image registration of multi-source remote sensing image is proposed.Firstly,the method of Maximally Stable Extremal Regions is adopted to detect the effective feature regions,the number of corners in each region is determined by the gray variance,subsequently,Harris is applied to extract the final feature points,ensuring the distribution of feature points is uniformly and non-redundant in the whole image.Secondly, remote sensing images are transformed by Contourlet.We construct multi-scale Gaussian combined moments on low frequency sub bands for initial matching,and then multi-direction Gray level co-occurrence moments are produced on high frequency sub bands for fine image registration.Lastly the correct matching pairs are completed.Experiments are implemented on several multi-source image sets.The results show that the proposed algorithm has advantages in terms of the quantity of points,control point distribution and registration accuracy,which can provide reference for future research.
作者
刘欢
肖根福
欧阳春娟
谭云兰
LIU Huan;XIAO Genfu;OUYANG Chunjuan;TAN Yunlan(Institute of Electronic and Information Engineering,Jinggangshan University,Ji'an,Jiangxi 343009,China;Key Laboratory of Watershed Ecology and Geographical Environment Monitoring, National Administration of Surveying,Mapping and Geoinformation,Ji'an,Jiangxi 343009,China)
出处
《遥感信息》
CSCD
北大核心
2018年第6期61-70,共10页
Remote Sensing Information
基金
国家自然科学基金(61462046
61640412)
江西省科技厅自然科学基金(20161BAB202049)
国家测绘地理信息局重点实验室招标项目(WE2016013
WE2016015)
江西省教育厅科技项目(GJJ160741
GJJ170633)
江西省高校人文社科项目(YS1546)
2016年度江西省艺术科学规划项目(YG2016250)
关键词
初-精结合
MSER-Harris
多尺度高斯组合矩
多方向灰度共生矩
多特征融合
图像配准
initial-fine combination
MSER-Harris
multi-scale Gaussian combined moment
multi-direction gray level co-occurrence moment
multi-feature fusion
image registration
