SAR and optical remote sensing image registration method based on pseudo twin convolutional neural network

Abstract

The invention discloses an SAR and optical remote sensing image registration method based on a pseudo twin convolutional neural network in the technical field of remote sensing image registration. Firstly, feature image blocks are collected and matched; abnormal point removal and final registration are carried out, a strategy of maximizing a feature distance between a positive sample and a hard negative sample is adopted, a new loss function is defined to train the network, and two branches of the pseudo twin network are connected through convolution operation to obtain a similarity score between two input image blocks. According to the invention, a pseudo twin convolutional neural network system structure is provided, so that the left branch and the right branch of the pseudo twin networkcan respectively input optical and SAR remote sensing images with different sizes, and the task of identifying corresponding image blocks in the optical and SAR remote sensing images under extremelyhigh resolution can be solved.

Description

technical field [0001] The invention relates to the technical field of remote sensing image registration, in particular to a SAR and optical remote sensing image registration method based on a pseudo-twin convolutional neural network. Background technique [0002] In recent decades, earth monitoring through Remote Sensing (RS) has been widely used in both military and civilian life. Multimodal remote sensing images contain a lot of complementary information, which is beneficial to many remote sensing applications. To this end, image registration is a common requirement to exploit multimodal images. However, due to the different imaging mechanisms, multimodal image registration is more challenging than ordinary image registration, especially for optical and Synthetic Aperture Radar (SAR) images. Finding common features in optical and SAR images is very difficult due to the different imaging mechanisms. And as the spatial resolution increases, the existing gap in geometrica...

AI Technical Summary

Problems solved by technology

[0008] The purpose of the present invention is to provide a SAR and optical remote sensing image registration method based on a pseudo-twin convolutional neural network to solve the problems of poor registration effect and low precision between SAR and optical images proposed in the above-mentioned background technology

Images