A panchromatic sharpening method of remote sensing image based on convolution neural network

Abstract

The invention discloses a panchromatic sharpening method of remote sensing image based on convolution neural network, includes the following steps: reading the original multispectral remote sensing image and its matching panchromatic remote sensing image; preprocessing the image data to obtain training sample; constructing convolution neural network structure; the training samples being inputted into the convolution neural network, and the loss function being stabilized at the minimum value by using the stochastic gradient descent algorithm, thus the optimal solution of the network structure is obtained. The same preprocessed test samples are input into the optimal convolution neural network structure, and the output is processed to obtain high-resolution multi-spectral remote sensing images. The invention can effectively reduce the spectral distortion of the processing result and enhance the sharpening effect.

Description

technical field [0001] The invention relates to the field of remote sensing image processing, in particular to a method for panchromatic sharpening of remote sensing images based on a convolutional neural network. Background technique [0002] Remote sensing images are one of the image types that are widely concerned at present. They are widely used in agricultural development, environmental monitoring, geological monitoring and other fields, and have good engineering application value and prospects. However, in practical applications, remote sensing images with high spatial resolution and spectral resolution cannot be acquired simultaneously due to the limitation of the physical structure of the sensor. In order to solve this problem, satellites generally have two different types of sensors, which respectively acquire panchromatic remote sensing images with high spatial resolution and multispectral remote sensing images with high spectral resolution. By using panchromatic ...

AI Technical Summary

Problems solved by technology

However, since multispectral remote sensing images have rich spatial information and spectral information, directly using convolutional neural networks to learn the mapping relationship between low-resolution and high-resolution multispectral remote sensing images will not only greatly increase the training time of convolutional neural networks , and it will also increase the learning error

Images