基于改进SIFT算法的多源遥感影像配准研究

Research on the multi-source remote sensing image registration based on the improved algorithm of SIFT

提出了一种新的基于SIFT(scale invari-ant features transform)算法的遥感图像配准算法。首先用Harris角点检测法代替了传统SIFT算法中的图像特征检测方法,再采用SIFT算法中的特征描述对图像特征进行描述,最后在参考图像上确定搜索范围,利用两幅图像之间的相关系数建立一对一的匹配关系。该算法不仅在特征点提取时间上进行了优化,同时避免了图像特征分布均匀时匹配效率低的问题。运用此算法对遥感图像进行配准实验,结果表明,该算法在保证匹配精度的同时有效地提高了匹配速度。

A new algorithm based on SIFT is proposed for remote sensing image registration. Firstly,the traditional image feature detection in SIFT algorithm is replaced by the Harris comer detection algorithm. Then we use SIFT algorithm to describe the image feature, finally the search range in reference images is determined and the one-on-one matching relation is established with the correlation coefficient of the images. The algorithm optimizes the feature point extracting time of traditional SIFT algorithm and avoids the low matching efficiency problem when the image characteristics is uniformly distributed. The matching experiment to SPOT satellite images is implemented with the new algorithm, and the results show that the new algorithm can enhance the matching speed effectively as the matching precision is ensured.