Research on Remote Sensing Image Registration Algorithm Based on Saliency Mechanism
Jiahao Wang, Yin Chen, Zhang Wencheng · 2022
In order to solve the problem that remote sensing image registration is easily affected by resolution, viewing angle, weather and other problems, resulting in low real-time and poor accuracy of registration results, a remote sensing image registration algorithm based on improved network in network is proposed. Use channel attention to find important channels and filter out useful feature information; design multi-scale matching at the same time, keep fewer parameters, speed up convergence, and improve robustness; improve the loss function, and use the four parameters obtained by regression to design a new loss The function iteratively optimizes the network model to improve the registration accuracy. The experimental results show that the algorithm has higher registration accuracy and faster running efficiency.