A Two-Stage Line Matching Method for Multi-temporal Remote Sensing Images
Zhengbing Wang, Jianhua Nie, Dan Li, Xugang Feng, Yuxiu Wu, Guili Xu · 2021 IEEE 6th International Conference on Signal and Image Processing (ICSIP) · 2021
Robust feature matching in multi-temporal remote sensing images is an essential but difficult task due to certain extent of appearance difference. In this paper, a two-stage line matching method is presented to match salient structures in multi-temporal remote sensing images. The proposed method consists of three main steps. In the first step, we extract the line structures from two corresponding images using the EDLines algorithm, and a line validation procedure is then implemented to remove meaningless and fragmented lines. In the second step, the detected lines are further divided into a set of line segments by a length threshold, and a new descriptor is computed to describe each line segment. After that, a coarse matching process is implemented to compute the projective transformation model. Finally, we adjust the length threshold according to the transformation model, and recompute the descriptors to further refine the matching process. We conduct the experiments on two pairs of multi-temporal remote sensing images, and the results demonstrate that the proposed approach can achieve favorable performance compared to common image matching methods.