Remote sensing image registration based on dual-channel neural network and robust point set registration algorithm

Abstract

Remote sensing image registration technology has important applications in military and civilian fields such as ground target recognition, urban development evaluation, and geographic change evaluation. In this paper, a remote sensing image registration method based on dual-channel convolutional neural network (DCCNN) is proposed. Firstly, the dual-channel neural network model with improved dense structure is used to extract the features of the input image pair and generate the corresponding feature points. Then the affine transformation coefficient is obtained by feature matching using the robust point set registration algorithm (TPS-RPM) based on thin-plate spline. Finally, the image to be registered can be transformed according to the coefficient to achieve the purpose of registration.The experimental results show that the registration accuracy of this method is higher than that of the comparison method.