摘要
提出一种新的异源高分辨率光学卫星遥感影像自动匹配方法。首先利用高分辨率光学遥感影像有理多项式模型(RFM)正解模型,通过迭代计算其反解模型,并将RFM反解模型与投影轨迹法结合提取特征点对应异源影像核线,通过核线检查与对比分析确定同名点搜索范围。然后在搜索范围内利用最小欧式距离准则提取初始同名点,最后采用RANSAC算法和多项式拟合迭代法剔除误匹配点以获取最终的匹配结果。试验结果表明,相对于传统的SIFT算法,本文方法通过对异源影像特征点构建核线几何约束和灰度分析进行匹配,可获取更好的匹配效率和精度。
A new matching method is presented for multi-source high resolution optical satellite images.In this method,an inverse RFM is iteratively built by using positive RFM,the areas in which corresponding points are searched are secondly delimited by the inverse RFM and projection tracking method.Then the minimum Euclidean distance criterion is used to extract the initial homonymy points within a given search area.Last,RANSAC and iterative polynomial error checker are imbedded to eliminate mismatching points.Experimental results verify that the presented method,which searches for control point based on geometric constraint of epipolar line and gray distribution,has higher matching rate than that resulted from SIFT matching method.
出处
《测绘学报》
EI
CSCD
北大核心
2013年第1期80-86,共7页
Acta Geodaetica et Cartographica Sinica
基金
国家自然科学基金(40771159
41101452)
关键词
异源
匹配
RFM
高分辨率光学遥感影像
核线
multi-source
matching
RFM
high resolution optical satellite images
epipolar