SAR image registration method, equipment and medium based on multi-feature constraints

Abstract

Embodiments of the present invention disclose a method, device and medium for SAR image registration based on multi-feature constraints. The method includes: extracting a ratio gradient map and a phase consistency feature map of a reference image and an image to be registered; determining the reference image and the to-be-registered image according to the ratio gradient map and the phase consistency feature map key points; through the trained twin dense connection network, extract the depth features of the neighborhood image blocks around the key points of the reference image and the image to be registered respectively; according to the matching situation of the depth features, predict the image matching A registration result, wherein the image registration result includes: matching keypoint pairs of the reference image and the to-be-registered image. This embodiment improves the registration accuracy and registration efficiency of the SAR image.

Description

technical field [0001] Embodiments of the present invention relate to the technical field of image registration, and in particular, to a method, device and medium for SAR image registration based on multi-feature constraints. Background technique [0002] Synthetic Aperture Radar (SAR) is an active microwave sensor whose imaging is not affected by light and weather, and has high light and seasonal stability. In the process of SAR image processing, it is often necessary to analyze and compare multiple images obtained in different periods of the same target area, such as change detection, information fusion, 3D reconstruction and map correction. A prerequisite for such tasks is to transform multiple images into the same coordinate system. Taking one of the images as a reference image and establishing a relative coordinate system, the other images to be registered are transformed into the coordinate system where the reference image is located through coordinate transformation....

AI Technical Summary

Problems solved by technology

[0004] However, due to the influence of large dynamic range and multiplicative noise, and the SAR images collected from different angles at different times are quite different, especially the SAR images obtained from different SAR sensors will have different geometric and radiation distortions. As a result, it is difficult to extract key points from the reference image and the image to be registered, the stability of the extracted features is poor, the number of effective features extracted is small, and the feature matching is difficult, which leads to the failure of the final SAR image registration or the registration accuracy is low. Difference

Images