Structural information guided cross-domain image geometric registration method

Abstract

The invention discloses a structural information guided cross-domain image geometric registration method, which comprises the following steps: acquiring a source image and a target image shot from different angles for the same region, constructing a cross-domain image geometric registration network for the two images, and performing feature extraction guided by image structural information; and performing cross-domain image geometric registration network training to form a cross-domain image geometric registration network model, sending the source image and the target image into the trained cross-domain image geometric registration network model to obtain geometric transformation parameters between the source image and the target image, and performing geometric transformation and pixel interpolation on the source image according to the geometric transformation parameters to obtain the target image, so that the source image and the target image are located in the same coordinate system, and then global registration of the cross-domain image is completed. According to the invention, the structure information of the image pair is utilized to guide network training, so that the influence of cross-domain image feature difference is reduced, and the accuracy of cross-domain registration is improved.

Description

technical field [0001] The invention relates to a cross-domain image geometric registration method guided by structural information. Background technique [0002] Image registration is the process of aligning two or more images of the same target taken at different times, from different angles, or from different sensors. Given a source image and a target image, due to different image imaging conditions, the image may have great differences in illumination, contrast, scale, rotation, grayscale change, etc. The main purpose of image registration is to overcome these difficulties and find a The optimal spatial geometric transformation from the source image to the target image, transforming two or more images into the same coordinate system, so that the transformed source image has the greatest similarity with the target image relative to a certain similarity metric function . [0003] For image registration tasks, it generally includes three steps: feature extraction, feature...

AI Technical Summary

Problems solved by technology

[0005] In the existing image registration method, the "Heterogeneous Image Matching Method Based on Deep Learning" with the publication number CN108537264A sends the heterogeneous image pair with the label (0,1) into the VGG network to obtain the feature map, and the feature map is differentiated. Alternately train the network after fusion and normalization, and convert image registration into a simple binary classification problem; Publication No. CN109711444A "A New Method for Remote Sensing Image Registration Based on Deep Learning" trains the convolutional neural network from feature points Obtain feature vectors, feature matching and transformation model estimation are all using traditional methods; Publication No. CN108346162A "Remote Sensing Image Registration Method Based on Structural Information and Spatial Constraints" uses structural information to generate feature descriptors, and uses spatial constraints to filter matching point pairs, The image configuration is completed based on the matching point pairs, and the calculation efficiency of the traditional calculation method is low, and the problem of cross-domain image registration cannot be solved.

Images