A cloud detection method based on sentinel-2 aerosol band

Abstract

The invention discloses a cloud detection method based on Sentinel-2 aerosol band, which is characterized by comprising the following steps: 1) image collection; 2) generating minimum image; 3) obtaining the mean value and standard deviation of the minimum image 4) Obtain the lower limit of segmentation threshold; 5) Generate mean image; 6) Obtain the mean and standard deviation of the mean image; 7) Obtain the upper limit of segmentation threshold; 8) Calculate segmentation threshold; 9) Image cloud detection. This method is less restricted by objective factors. It can show robust effects for the identification of thick, cirrus, and thin clouds in images, and can also provide better quality for subsequent tasks such as atmospheric parameter inversion, image registration, and fusion. It has very important practical value for the research of large-scale land use mapping and crop yield estimation.

Description

technical field [0001] The invention relates to the fields of remote sensing technology and cloud microphysics, in particular to a cloud detection method based on the Sentinel-2 aerosol band. Background technique [0002] Remote sensing images are widely used in large-scale remote sensing land use mapping, crop growth monitoring, crop yield prediction, and pest monitoring. With the continuous sharing of high / higher resolution satellite data, the application value of remote sensing images is becoming more and more important in the study of high-precision Land-Use and Land-Cover Change (LUCC). However, optical remote sensing images are extremely susceptible to cloud noise, resulting in many data default areas in images, which poses a huge challenge to the interpretation accuracy and reliability of large-scale LUCC research based on high / higher resolution remote sensing. [0003] Current research related to cloud detection mainly includes reflectance threshold method, texture ...

AI Technical Summary

Problems solved by technology

The albedo threshold method is easily affected by cloud noise edge mixed pixels; the texture analysis method is difficult to identify cirrus and thin clouds with complex textures; the multi-feature synergy method and the method of identifying cloud noise combined with machine can improve the accuracy of cloud detection, but there are still some misjudgments and missed judgments.

In addition, for Sentinel-2 remote sensing images with high spatio-temporal resolution, cloud detection based on the QA60 band is difficult to obtain satisfactory recognition results; some studies based on three near-infrared bands of Sentinel-2 images for cloud detection have obtained comparative High overall accuracy, but cloud and snow aliasing at high altitudes

Images