Abstract: This paper proposes a robust and accurate algorithm based on SURF descriptor for multi-source remote-sensing images registration. The algorithm detects the matching points by SURF. A large number of error matching points are eliminated by linear polynomial transformation, Forstner detector and least square with coarse to fine registration process. The image is divided to be corrected accurately through constructing the Triangulated Irregular Network(TIN) with dense refined matching points. Experimental results show the algorithm can realize precise registration for multi-source remote sensing images.

Key words: image registration, SRUF descriptor, Forstner detector, Triangulated Irregular Network(TIN)

摘要： 针对多源遥感影像的配准问题，提出一种基于SURF描述子的匹配算法，该算法具有稳健及高精度特性，采用SURF匹配算法得到同名点，利用一次多项式、Forstner算子和最小二乘原理由粗及精地剔除残差较大的同名点，利用提取的密集同名点构建三角网，对影像进行分块精确纠正。实验结果表明，该算法能够实现多源遥感影像的高精度配准。

关键词: 影像配准, SURF描述子, Forstner算子, 三角网

CLC Number: 

• TP751.1