Multispectral cloud detection method based on semi-supervised spatial spectral features

Abstract

The invention discloses a multi-spectral image cloud detection method based on semi-supervised spatial spectrum features, which mainly solves the problem in the prior art that there are many false detections and missed detections of cloud targets in multi-spectral images with complex backgrounds and thick clouds. Realization of the present invention: input multi-spectral image; implement band selection on multi-spectral image, select visible light band; obtain the spatial feature image of multi-spectral image; extract cloud target spectrum in the spatial feature image; use constrained energy minimization algorithm to obtain multiple Cloud object detection results for spectral images. The present invention uses band selection to select images in visible light bands, and uses an image decomposition method to extract spectral features and spatial features for cloud target detection, which not only reduces the false detection and missed detection of thicker clouds, but also can better distinguish objects in multi-spectral images. Cloud targets and complex backgrounds have the advantages of fewer false detections of cloud targets and fewer missed detections of cloud targets in detection results, and are applied to cloud detection in remote sensing images.

Description

technical field [0001] The invention belongs to the technical field of image processing and relates to cloud detection of remote sensing images, in particular to a multispectral cloud detection method based on semi-supervised spatial spectrum features. The invention can be used to detect cloud targets from remote sensing images. Background technique [0002] With the rapid development of remote sensing technology, remote sensing images are widely used in various fields such as military target recognition, environmental monitoring, meteorological analysis, mineral development, geographic surveying and mapping, etc. However, not all images can meet the requirements of intelligent information processing, and one of them is An important factor is cloud cover, with statistics showing that at any one time 50% of the Earth's surface is covered by clouds. The existence of clouds, on the one hand, has blocked the ground features, resulting in the lack of surface information, which h...

AI Technical Summary

Problems solved by technology

Although this method can improve the accuracy of cloud detection by constructing a Gaussian mixture model, the disadvantage of this method is that due to the complex background of multispectral images and a lot of interference information, the cloud detection method can be directly applied to complex background and Multispectral images with thick cloud layers are prone to false detection and missed detection, resulting in low accuracy of cloud detection

Although this method uses the M-type convolution model to achieve pixel-level semantic segmentation, and finally achieves a better cloud detection effect, the disadvantage of this method is that due to the large number of bands in the spectral image, the M-type convolution The training process of the product network is complex and cumbersome, and the use of the pixel-level cloud region detection method will cause holes and breaks in the detection results, resulting in missed detection results, increased false alarm rates, and reduced cloud detection accuracy.

[0005] Existing cloud detection technologies rely heavily on thermal infrared spectroscopy and are mostly pixel-level cloud detection methods. They require too high resolution of remote sensing images, and the cloud detection accuracy of multispectral images with complex backgrounds and thick clouds is low.

Images