A method and system for removing thick clouds from remote sensing images

Abstract

The invention discloses a method and system for removing thick clouds from remote sensing images. The method includes: dividing the target image into a cloud area image and a cloud-free area image; clustering the pixels in the cloud-free area image to obtain a pixel category group; searching for similarity pixels between the reference image and the cloud-free area image; calculating each The proportion of class similarity pixels to the pixels in the pixel category group; if there is at least one class whose proportion is greater than the first preset proportion, the similarity pixel with the largest proportion is determined as the cloud-free similarity pixel; if all categories The ratios of which are all smaller than the first preset ratio, determine the category of the similarity pixels whose ratio is greater than the second preset ratio; count the number of pixels of each category in the cloud mask buffer; determine the corresponding category of the largest number of pixels The similarity cells of are the declouded similarity cells; replace the cells in the cloud region image with the declouded similarity cells. The invention can eliminate the influence of the change of the ground surface on the two-phase images on the judgment of similarity pixels, and improve the accuracy of thick cloud removal.

Description

technical field [0001] The invention relates to the technical field of remote sensing image processing, in particular to a method and system for removing thick clouds from remote sensing images. Background technique [0002] In recent years, remote sensing has been widely used, ranging from land and ocean monitoring to inversion of related parameters. For any remote sensing application, it is very important to obtain valid data for the desired area. Remote sensing images are often blocked by clouds and cannot detect the ground, making it difficult to obtain information on the earth's surface, which severely limits the large-scale and long-term observation of remote sensing technology. Therefore, although a large number of remote sensing images can be obtained, the available "cloud-free and effective" image data may not meet the actual needs, so how to obtain high-quality cloud-free remote sensing data in a specific time and region is still an urgent problem to be solved. q...

AI Technical Summary

Problems solved by technology

The NSPI method uses image mosaic and stitching technology to mosaic the cloud-free part into the "cloud pollution" area to obtain a cloud-free image. This method uses the data of the cloud-free area of ​​the reference image to fill the cloudy area of ​​the target image. When the multi-temporal data When the time difference between the two phases of images is not large and the change of the surface characteristics is not obvious enough, this method can achieve a good cloud removal effect. Spectral differences, which will affect the judgment of similarity pixels, introduce a large number of non-similarity pixels, resulting in low accuracy of thick cloud removal

Images