An accurate registration method of optical remote sensing images based on vgg network and Gaussian difference network

Abstract

The invention provides an optical remote sensing image accurate registration method based on a VGG network and a Gaussian difference network. First, a Gaussian difference network of an input image is constructed, and a maximum point is extracted as a candidate key point, and the coordinates and the scale of the key point are obtained; Then resize the image and input it into the pre-trained VGG network to obtain feature maps of different sizes; stitch the Gaussian difference images and feature maps of the same size to form a DVGG network; extract the VGG using the coordinates of the candidate key points The feature vector of the network is spliced ​​with the corresponding SIFT feature vector to form the output vector of the DVGG network, which is used as the feature description of the key points; the similarity between the key points in the image to be registered and the reference image is calculated, and the matching pairs are obtained. Calculate The transformation matrix is ​​obtained to obtain the registration image. Compared with the existing registration method, the present invention obtains more correct matching point pairs, and the registration effect is better.

Description

technical field [0001] The invention relates to the technical field of digital image processing, in particular to an optical remote sensing image registration method combining VGG and Gaussian difference network. Background technique [0002] Image registration technology refers to the process of matching two images acquired at different times, different perspectives and different sensors. It has been widely used in remote sensing, medicine and other fields. Among them, image registration in the field of remote sensing is an indispensable part of many remote sensing image processing processes. For example, tasks such as target detection, image fusion, and semantic segmentation need to be carried out on the premise of successful registration. Therefore, the registration technology of optical remote sensing images It has always been a research hotspot. [0003] The traditional image registration methods are mainly divided into gray-based methods and feature-based methods. The...

AI Technical Summary

Problems solved by technology

However, the current image registration method based on deep learning is mainly applied to natural images. Remote sensing images usually have higher resolution and larger size, the number of key points extracted by CNN is small and the accuracy is poor, and the registration accuracy is low.

Images