Panchromatic and multispectral image fusion method and device and application thereof

Abstract

The invention discloses a panchromatic and multispectral image fusion method and device and application thereof, and the method comprises the steps: carrying out the wave band superposition of a panchromatic image, carrying out the interpolation of a multispectral image into an interpolation image with the same resolution as the panchromatic image, carrying out the subtraction of the two images, and carrying out the Gaussian filtering, and obtaining an interpolation trend chart; obtaining a fusion model with high-fidelity fusion capability by constructing an unsupervised fusion model with a spectrum discriminator and a spatial spectrum discriminator; and automatically processing the panchromatic and multispectral images by using the fusion model to obtain a high-fidelity fusion image. The fusion model obtained through the fusion method is suitable for processing the remote sensing image with the original resolution, the common detail distortion problem of a current fusion method based on deep learning is effectively reduced, the operation speed is high, and the fusion effect is good.

Description

technical field [0001] The invention relates to the technical field of satellite remote sensing image fusion, and more specifically relates to a panchromatic and multispectral image fusion method, device and application thereof. Background technique [0002] With the rapid development of satellite remote sensing technology, more and more multi-sensor satellites are launched into space, such as IKONOS, QuickBird, Zi Yuan, WordView-2 and other satellites. Satellite remote sensing technology has become an important indicator to measure the comprehensive strength of countries. A large amount of high-resolution and multi-channel satellite remote sensing data produced by satellites can provide information services and decision support for important fields such as modern agriculture and ecological construction. [0003] In the practical application of satellite remote sensing images, people often hope that the remote sensing images acquired by satellites have high spectral and spat...

AI Technical Summary

Problems solved by technology

This type of algorithm benefits from the excellent nonlinear fitting characteristics of the deep neural network, the fusion effect is stable, and it has outstanding performance in spectral fidelity. However, the spatial detail information of this type of algorithm is not fully preserved, resulting in low spatial fidelity of the fusion result.

For a long time, researchers have been committed to improving the network model and increasing the spatial detail information in the fusion results, but there are still two problems that have not been resolved: 1) Most of the existing deep learning-based fusion methods downsample the multispectral and panchromatic images , and then use the original multispectral image as the truth map, the parameters obtained by training are difficult to generate remote sensing images at high resolution

2) In order to improve the spatial detail information, the existing deep learning-based method directly adds the upsampled panchromatic image and the detailed information output by the network to obtain the final fusion result. The upsampling operation is easy to introduce noise and it is difficult to optimize the network parameters. is improved, thus causing distortion of spatial details locally in the fused image

Images