Satellite image automatic cloud detection method based on Gaussian mixture model

Abstract

The invention discloses a satellite image automatic cloud detection method based on a Gaussian mixture model. The satellite image automatic cloud detection method comprises a step S1 of conducting statistics on raw grayscale histograms of satellite images, and pre-processing the grayscale histograms to eliminate interference; a step S2 of constructing a Gaussian mixture module for the grayscale histograms; a step S3 of determining a grayscale threshold for segmenting a cloud region; and a step S4 of segmenting the cloud region using the grayscale threshold. The method of the invention makes up for the shortcomings of a traditional texture analysis method, a homomorphic filtering method and a multi-spectral synthesis method, and can be used for satellite image screening and quality control, cloud zone automatic extraction, cloud-containing image quick repair and so forth. The invention is based on single-band image histograms, not limited to a satellite spectral range, and suitable for both panchromatic and multispectral images; and is high in detection accuracy, requires no supplementary information and manual intervention, and is high in calculation speed and capable of automated processing.

Description

technical field [0001] The invention belongs to the technical field of remote sensing image processing, in particular to an automatic cloud detection method for satellite images based on a Gaussian mixture model. Background technique [0002] Represented by the WorldView series, Pleiades series, and China's "Gaofen" series of satellites, high-resolution earth observation technology has developed rapidly, satellite revisit cycles have been continuously shortened, satellite image resolution and width have been continuously improved, and the data volume of satellite images It is widely used in resource investigation, disaster prevention and mitigation and other fields. However, in the process of satellite image data processing, the existence of clouds is a problem that cannot be ignored. It not only blocks the ground scene, but also changes the spectral characteristics of the image to a certain extent, causing many problems for image matching, color uniformity and mosaic work. ...

AI Technical Summary

Problems solved by technology

The multi-spectral synthesis method uses the difference in cloud and clear sky reflectivity, but it requires the sensor to be equipped with multiple thermal infrared bands, and the detection wavelength needs to cover the absorption band of water or carbon dioxide, while the earth observation satellite image generally only contains blue, green, red , near-infrared 4 bands, the spectral range is generally 0.43 ~ 0.90 μm, the advantages of multi-spectral synthesis method is difficult to reflect

Images