Improved attention complementary dual channel network remote sensing image registration
Wei Wang, Ying Chen, Zhang Wencheng, Jiahao Wang, Xianjing Li, Qi Zhang · 2022
An end-to-end registration algorithm based on deep learning is proposed to solve the problem of low accuracy and low efficiency of existing remote sensing image registration methods. Firstly, in the feature extraction stage, a double channel feature extraction network is proposed based on the principle of multi visual feature complementarity and the attention module, which makes the network pay more attention to the important channels and spaces of the image. Finally, in the matching phase, the feature relationship between the two directions is obtained by bidirectional correlation feature matching, which avoids the inaccuracy of single direction matching. Experimental results show that the proposed algorithm outperforms other algorithms on multiple datasets, and effectively improves the accuracy and efficiency of remote sensing image registration.