论文部分内容阅读
目的多视角遥感图像配准是遥感图像处理领域的一项关键技术,其目的是精确获取图像间被测区域发生的几何变换关系。由于航拍视角变化以及地物的空间分布和几何形状的复杂度,多视角遥感图像间会产生非刚性畸变问题,增加了图像配准的难度,为此本文提出一种利用遥感图像SIFT(scale-invariant feature transform)特征点阵的全局和局部几何结构特征进行多视角遥感图像配准的算法。方法通过增加对SIFT点阵的几何结构特征描述以及利用SIFT点阵间全局与局部几何结构特征的互补关系,提升存在非刚性畸变的多视角遥感图像配准精度。结果实验使用谷歌地球的卫星影像数据以及无人机航拍遥感数据对本文算法进行了测试,并与3种同类算法(SIFT、SURF(speeded-up robust features)、CPD(coherent point drift))进行对比实验,本文算法在存在非刚性畸变的多视角遥感图像配准中能够有效地提升SIFT特征点阵的配准精度,从而获得更加准确的多视角遥感图像配准结果。结论本文实现了一种结合SIFT特征点阵的全局和局部几何结构特征进行多视角遥感图像配准的算法,实验结果表明了该方法对存在非刚性畸变的多视角遥感图像能够有效地进行配准,可适用于同源多视角情况下的遥感图像配准问题。
Purpose Multi-view remote sensing image registration is a key technology in the field of remote sensing image processing. Its purpose is to accurately obtain the geometric transformation relationship of the measured area between images. Due to the variation of the aerial view and the spatial distribution and the complexity of the geometric features, the non-rigid distortion problem occurs in multi-view remote sensing images and the difficulty of image registration is increased. Therefore, invariant feature transform The global and local geometric features of the feature lattice An algorithm for multi-view remote sensing image registration. The method improves the registration accuracy of multi-view remote sensing images with non-rigid distortion by increasing the description of the geometric features of the SIFT lattice and using the complementary relationship between the global and local geometric features of the SIFT lattice. Results Experiments using Google Earth satellite imagery and UAV aerial remote sensing data have tested the algorithm and compared it with three kinds of similar algorithms (SIFT, SURF (speeded-up robust features), CPD (coherent point drift)) Experiments show that the algorithm proposed in this paper can effectively improve the registration accuracy of SIFT feature points in multi-view remote sensing image registration with non-rigid distortion so as to obtain more accurate multi-view remote sensing image registration results. Conclusions In this paper, an algorithm for multi-view remote sensing image registration based on the global and local geometric features of the SIFT feature lattice is implemented. The experimental results show that this method can efficiently register multi-view remote sensing images with non-rigid distortion , Which can be applied to the registration of remote sensing images with the same source and multiple views.