Fusion method suitable for high-resolution remote sensing image

Abstract

The invention relates to a fusion method suitable for high-resolution remote sensing images. The invention relates to the field of remote sensing image fusion, in particular to a design of a fusion method suitable for a high-resolution remote sensing image. The method mainly solves the problem that a high-resolution panchromatic image and a multispectral image with a slightly lower resolution are fused to obtain a high-resolution multispectral image. According to the invention, a YOLACT deep learning method is utilized to extract buildings in a high-resolution panchromatic image, a sliding window is intercepted from the image, a building weight is obtained through calculation according to the number and distribution of the buildings in the sliding window, then on the basis of an SFIM fusion method, parameters of fusion operation are adjusted in a self-adaptive mode along with changes of the building weight, and therefore, the fusion effect of the building area and the non-building area is optimal, and finally a high-resolution multispectral image with better visual effect and definition and richer information amount is obtained.

Description

technical field [0001] The invention relates to the technical field of remote sensing image fusion, and is a fusion method suitable for high-resolution remote sensing images. Background technique [0002] It is often difficult to take both high-resolution and multi-spectral into account when acquiring remote sensing images, so the technology of fusing high-resolution panchromatic images with slightly lower-resolution multi-spectral images to obtain high-resolution multi-spectral images is very important . Remote sensing image fusion technology can complement the advantages of the above two images, provide users with richer information, more realistic and clear fusion images, and help improve image feature extraction, target classification and recognition accuracy, and enhance data practicability. [0003] At present, the methods of remote sensing image fusion are mainly divided into three categories: pixel-level fusion, feature-level fusion and decision-level fusion. The c...

AI Technical Summary

Problems solved by technology

[0005] IHS transform and Brovey transform only support three bands, while most multispectral images currently have at least four bands (blue band, green band, red band, and near-infrared band), so the use scenarios of these two methods are limited; PCA Because the information of the first component of the transformation method is too concentrated, the fused image will have a large spectral distortion (that is, color distortion); the GS transformation can improve the problem of spectral distortion to a certain extent, but the calculation is large, and it is not suitable for processing large high resolution remote sensing imagery

The PanSharp method is affected by the extreme grayscale area, which can easily lead to color distortion of the fusion result; the fusion result of the traditional SFIM method has low spatial information integration and poor clarity; the improved SFIM method calculates the average gradient for MS, and the spatial feature information of the building area The spatial feature information of non-building areas (such as vegetation, bare soil, etc.) is relatively sparse and the average gradient is small. However, this method uses the same set of fusion parameters for the above-mentioned areas, resulting in It is difficult to obtain the best results for both non-building areas and non-building areas, and the visual effect, clarity and information content of the fusion image are still not ideal

Images