尺度不变特征转换算法的多源SAR影像匹配

Multi-sensor SAR Imagery Matching by Using SIFT

鉴于直接利用SIFT算法进行SAR影像间的匹配不能得到很好的效果,考虑SIFT算法在应对噪声以及对镜像影像进行匹配的局限性,提出了针对SAR影像之间匹配的SIFT算法预处理。首先利用影像与影像之间的空间信息进行匹配,之后利用SIFT算法进行局部特征点匹配,通过采用RANSAC进行错配点的去除,实现SAR影像的高精度配准。实验结果表明,该文提出的预处理以及错配点的去除为利用SIFT算法进行SAR影像的匹配提供了可能。

Scale Invariant Feature Transform （SIFT）has been widely applied in image processing.To a certain extent,the algorithm can keep scale and rotation invariance and weaken the influence of the light.However,directly using SIFT for SAR image matching could not obtain good results as expected because of the speckle noises caused by SAR imaging.In consideration of the limitation of SIFT in noises and mirror image situation,this paper proposes a pre-processing method of SIFT algorithm for SAR image matching,which firstly match images using the geographic information between them,then match local feature points with SIFT and remove mismatching points with RANSAC,finally realize the high-accuracy matching of SAR images. Results show that the pre-processing method proposed by this paper and the removal of mismatching points provide the possibility for SAR image matching using SIFT.