论文部分内容阅读
针对存在云雾遮挡、仿射变形的遥感影像,本文提出应用Hessian-Affine与最大信息熵,检测并筛选仿射不变特征,同时选刺同名点估计初始变换参数,对每一个待匹配点预测出一定圆域约束内的对应匹配点集,利用NCC相关系数迭代确定圆域内真同名点,得到初始匹配点集,然后利用均方根误差(RMSE)迭代剔除误匹配,直至完成最佳仿射变换参数估计。结果表明:该算法对发生仿射畸变、气候复杂区域的影像配准表现出较好稳健性和定位精度。
Aiming at the remote sensing images with cloud occlusion and affine deformation, this paper proposes to use Hessian-Affine and maximum information entropy to detect and select affine invariant features and estimate the initial transformation parameters of the same name at the same point, and predict each matching point A set of corresponding matching points within a certain circular domain constraint, and use the NCC correlation coefficient to iteratively determine the true namespace in the circular domain to obtain the initial matching point set, and then reject the mismatch using the root mean square error (RMSE) iteration until the best affine transformation is completed Parameter Estimation. The results show that the algorithm shows good robustness and positioning accuracy for image registration with affine distortion and complicated climate.