基于SIFT特征的遥感影像自动配准

Automatic Registration of Remote Sensing Imagery Based on SIFT

提出了一种基于特征点的全自动高分辨率遥感影像配准方法,该方法采用对于尺度具有鲁棒性的SIFT算法进行特征点的提取与匹配,并通过RANSAC方法和双向匹配策略提高特征点的匹配精确度。最后利用同名特征点构建影像间的转换模型,实现高精度影像纠正与配准。实验结果表明,该算法具有较强的匹配能力和鲁棒性。

An automatic and high-precision registration of remote sensing imagery based on feature points is proposed. First a scale-invariant feature extracting algorithm SIFT is used for feature extraction and matching. Then RANSAC method and bidirectional image matching algorithm are used to improve the accuracy of matehing points. Finally, the exacted matching points are used to precisely compute the transformation relations and transformation model of the two images, automatically matching the images. The experiments show that the algorithm is of strong matching ability and robustness.