一种基于SIFT特征的快速逐层遥感图像配准方法

Fast Hierarchical Registration Method for Remote Sensing Image based on SIFT

当前SIFT特征分层配准方法中存在特征点匹配复杂度高以及不同时相地物变化导致特征点误匹配等问题,提出一种基于SIFT特征的"低分辨率配准、高分辨率验证"快速逐层遥感图像配准方法。该方法针对同源同分辨率不同时相的遥感图像,通过在金字塔的低分辨率图层匹配特征点对并建立仿射变换模型,在金字塔的高分辨率图层评估并修正模型。实验表明:提出的方法在保证配准精度的前提下,有效提高了配准算法的效率。

For the current automatic image registration based on SIFT,feature point matching algorithm is time-consuming,in addition,the changes of multi-temporal images affect the accuracy of registration,this paper proposes a SIFT-based feature of the "low-resolution matching,high resolution authentication" hierarchical image registration algorithm to improve the above issues.In the proposed algorithm,affine transformation model is established in low-resolution pyramid images and sequentially evaluated and revised by match points in high resolution pyramid images.Experimental results show that the improved SIFT algorithm can reduce the time complexity with rather considerable accuracy.