基于MSER特征的无人机倾斜影像匹配算法

UAV image matching algorithm based on MSER

针对无人机影像存在仿射变形与阴影问题,文中提出应用MSER和影像最大信息熵,提取仿射不变特征,通过POS数据估计影像变换初始参数,利用初始变换参数估计待匹配点的位置,提取匹配点局部区域的点集,利用NCC相关系数确定局部区域内点的同名点,对于初始匹配点对采用均方根误差(RMSE)剔除初始中误匹配点对。实验结果表明,该算法在存在仿射变形、高大建筑物阴影的影像匹配表现较好的结果。

In order to solve the problem of affine transformation and shadowing in UAV images,in the paper,the affine invariant features was extracted based on the MSER and the maximum information entropy of the image. Initial parameters of the image transformation were estimated using POS data,the position of the point to be matched was estimated using the initial transformation parameters to extract the matching point locations. The same point in the interior of the point set was determined by using the NCC correlation coefficient,and the root mean square error（ RMSE） was used to eliminate the initial mismatched point pairs in the initial matching point set. Experimental results showed that the proposed algorithm performs were well in image matching with affine deformation and shadowing of tall buildings.