Remote sensing image automatic cloud detection method and device

Abstract

The invention discloses a remote sensing image automatic cloud detection method and device, and belongs to the technical field of satellite remote sensing. The method comprises the following steps: carrying out effectiveness judgment of pixel points, and primarily locking a cloud detection area; then, partitioning a remote sensing image, and respectively determining a detection threshold value of each wave band by utilizing mean values and variances of effective pixels in the image blocks in a near-infrared wave band, a short-wave infrared wave band and a thermal infrared wave band; and finally, performing preliminary cloud detection by using the detection threshold of the near-infrared band, removing a white object by using the detection threshold of the short-wave infrared band, and removing a false alarm by using the detection threshold of the thermal infrared band. According to the method, any parameter and labeled sample do not need to be manually set, the physical characteristic threshold is automatically selected by self-adapting to the spectral characteristics of the image, and one-button rapid, batch and high-precision cloud detection and cloud mask product production are realized; the calculation amount is reduced, and the cloud detection efficiency is improved; and artificial targets and highlight areas are eliminated, and misjudgment of cloud-like targets is reduced.

Description

technical field [0001] The invention relates to the technical field of satellite remote sensing, in particular to an automatic cloud detection method and device for remote sensing images. Background technique [0002] During the earth observation and imaging process of satellite remote sensing sensors, because the sensors are greatly affected by changes in atmospheric density and cloud cover, many images have the problem of cloud cover, which leads to attenuation of original ground object information, spectral distortion, and even loss, reducing the ground quality in remote sensing images. The availability of target information affects the interpretation and extraction of high-resolution remote sensing images, the inversion accuracy of quantitative products for ecological parameters, and the production of surveying and mapping products such as time-space cloudless remote sensing image mosaic products. Therefore, cloud detection is an important part of satellite remote sensin...

AI Technical Summary

Problems solved by technology

The disadvantage is that it is highly sensitive to the threshold, only applicable to local geographical areas or specific background areas or specific satellite sensors, and the universality is not strong

The texture analysis method mainly uses physical characteristics such as cloud shape, texture, and gray scale for cloud detection. The disadvantage is that the accuracy of cloud classification features is relatively high.

The frequency combination threshold method mainly uses the low-frequency characteristics of clouds, and obtains image low-frequency data through wavelet analysis, Fourier transform and other methods for cloud detection. The disadvantage is that the detection efficiency is low

Machine learning methods such as support vector machines need to obtain a large number of training samples, and have high requirements for the selection of classification features. They usually need to rely on auxiliary data to set thresholds, and the degree of businessization is low.

Images