Remote sensing image registration method based on visual transformer

The invention provides a remote sensing image registration method based on a visual transformer, and the method comprises the steps: carrying out the internal enhancement of a reference image in an input image at a preprocessing stage, and carrying out the affine transformation of the reference image through a spatial transformation network with priori knowledge; in the feature extraction stage, an algorithm combining a visual transformer network and a Shuffle network is provided, the lightweight visual transformer network learns a long-distance relation between image points by using sub attention, and the Shuffle network learns image spatial features and detailed positioning information to obtain high-dimensional feature information. In the feature matching stage, firstly, feature points are filtered by using a neighborhood consistent network, and then the similarity of corresponding features in two image blocks is calculated by using a cross-correlation algorithm. And finally, obtaining a registration resul