multispectral remote sensing image Pan-shift method based on a multilayer coupling convolutional neural network

Abstract

The invention discloses a multispectral remote sensing image Pan-based on a multilayer coupling convolutional neural network. According to the method, high-level features of an input image and an output image are extracted through two convolutional auto-encoders respectively, end-to-end connection is established between the input image and the output image of the network while a feature mapping layer is established between the two high-level features through the coupled convolutional network, and finally the overall network framework is finely adjusted. According to the method, the image LM-isfully extracted through an algorithm of the convolutional auto-encoder; HR and image HR- According to the method, the internal features of the MS are utilized, the convolutional neural network is established between the two internal features in a coupling convolution mode, and the connection is established between the input and the output of the network to form the end-to-end network, so that thefusion precision is improved.

Description

technical field [0001] The invention belongs to the field of image fusion, and in particular relates to a Pan-sharpening method based on a multi-layer coupled convolutional neural network. Background technique [0002] In order to understand the earth from a multi-dimensional and macro perspective, remote sensing has become an emerging technology. Remote sensing image is a kind of image that can detect and record various ground object information, mainly obtained by aerial photography or satellite photography. But for an optical remote sensor system, image spatial resolution and spectral resolution are a pair of contradictions. Under the condition of a given signal-to-noise ratio, higher spectral resolution often means that high spatial resolution cannot be achieved at the same time. To this end, a remote sensing image fusion technology, Pan-sharpening algorithm, is proposed. Specifically, the Pan-sharpening algorithm is a method that integrates the spatial details of a h...

AI Technical Summary

Problems solved by technology

Therefore, the self-encoder is not the most suitable method when extracting image features, which will affect the quality of the fusion image

At the same time, since the autoencoder is a fully connected layer network structure, there are too many network parameters, and the computer configuration requirements are too high, which is not suitable for image recognition tasks.

Since the gradient of the fully connected neural network cannot pass more than three layers, there is a limit on the number of layers of the network

However, the more layers of the network, the stronger its expressive ability, so the fully connected layer limits the expressive ability of the network.

Images