CN-121998843-A - Visible light remote sensing image thin cloud correction method, device, equipment and medium

CN121998843ACN 121998843 ACN121998843 ACN 121998843ACN-121998843-A

Abstract

The invention discloses a thin cloud correction method for visible light remote sensing images, which comprises the steps of obtaining a thin cloud image to be corrected and a cloud-free reference image, preprocessing the two images, constructing a differential domain based on the preprocessed thin cloud image to be corrected and the cloud-free reference image, performing independent component analysis on differential domain data to obtain a plurality of independent components and a mixed coefficient matrix, calculating the low-frequency energy duty ratio of each independent component, identifying the thin cloud independent component according to the low-frequency energy duty ratio, extracting a cross-band coefficient vector corresponding to the thin cloud independent component from the mixed coefficient matrix, and performing thin cloud stripping processing on the thin cloud image to be corrected based on the cross-band coefficient vector to obtain a cloud-free clear image after thin cloud correction. The invention effectively solves the problem of image blurring and ground surface information distortion caused by thin cloud, remarkably improves the definition and usability of the thin cloud image, and improves the quality and application value of the remote sensing image.

Inventors

ZHANG CHI
ZHU ZHAOHUAN
MEI LE
YE RUIMING
CHEN JIANHUA
MA LI
YI PENG
HU WENXIONG
LEI GUANGYUAN
GONG GUODONG
CHEN RUIXIAN

Assignees

广州市城市规划勘测设计研究院有限公司
广州花都规划勘测设计院有限公司

Dates

Publication Date: 20260508
Application Date: 20260409

Claims (10)

1. A thin cloud correction method for a visible light remote sensing image is characterized by comprising the following steps: acquiring a thin cloud image to be corrected and a cloud-free reference image, and preprocessing the two images; Constructing a differential domain based on the preprocessed thin cloud image to be corrected and the cloud-free reference image, and performing independent component analysis on differential domain data to obtain a plurality of independent components and a mixing coefficient matrix; calculating the low-frequency energy duty ratio of each independent component, identifying the thin cloud independent component according to the low-frequency energy duty ratio, and extracting a cross-band coefficient vector corresponding to the thin cloud independent component from the mixed coefficient matrix; and carrying out thin cloud stripping treatment on the thin cloud image to be corrected based on the cross-band coefficient vector to obtain a cloud-free clear image after thin cloud correction.
2. The method for correcting the thin cloud of the visible light remote sensing image according to claim 1, wherein the preprocessing of the two images comprises: Resampling the cloud-free reference image to enable the spatial resolution of the resampled cloud-free reference image to be consistent with that of the thin cloud image to be corrected; and carrying out band-by-band radiation unification treatment on the resampled cloudless reference image so as to ensure that the band of the uniformed cloudless reference image is consistent with that of the thin cloud image to be corrected.
3. The method for thin cloud correction of visible light remote sensing images according to claim 2, wherein the performing band-by-band radiation uniformization processing on the resampled cloud-free reference image comprises: According to the thin cloud image to be corrected and the resampled cloud-free reference image, calculating the total difference of each pixel in all wave bands; Selecting a stable pixel set based on the total difference, and calculating a radiation consistency parameter in a stable pixel area by adopting a least square method; And carrying out band-by-band correction on the resampled cloud-free reference image by using the radiation uniformization parameters to obtain the cloud-free reference image after radiation uniformization.
4. The method for correcting the thin cloud of the visible light remote sensing image according to claim 1, wherein the constructing a differential domain based on the preprocessed thin cloud image to be corrected and the cloud-free reference image, and performing independent component analysis on differential domain data to obtain a plurality of independent components and a mixing coefficient matrix comprises: Performing band-by-band difference on the preprocessed thin cloud image to be corrected and the cloud-free reference image to obtain a multiband difference image; Constructing an observation matrix of the multiband differential image, wherein each column of the observation matrix corresponds to a multiband differential vector of one pixel; and carrying out independent component analysis on the observation matrix to obtain three independent components and a mixing coefficient matrix, wherein the three independent components are respectively a thin cloud radiation disturbance, ground real change and a system residual error.
5. The method of claim 1, wherein the calculating the low frequency energy duty ratio of each independent component comprises: performing two-dimensional Fourier transform on each independent component image to obtain a corresponding frequency spectrum and calculating a power spectrum; Defining a low-frequency window by taking a frequency spectrum center as a reference, and counting the energy in the low-frequency window and the total energy of the whole frequency domain; The ratio of the low frequency energy to the total energy in the full frequency domain is taken as the low frequency energy ratio of the independent component.
6. The method of claim 1, wherein identifying the thin cloud independent component according to the low frequency energy duty ratio comprises: sequencing the three independent components from high to low according to the low frequency energy ratio; If the unique maximum low-frequency energy duty ratio component exists, determining the low-frequency energy duty ratio component as a thin cloud independent component; If a plurality of components with the largest low-frequency energy ratio are arranged in parallel, calculating the gradient energy of each independent component, and selecting the independent component with the smallest gradient energy as the thin cloud independent component.
7. The method for thin-cloud correction of visible light remote sensing images according to claim 1, wherein the thin-cloud stripping process is performed on the thin-cloud image to be corrected based on the cross-band coefficient vector to obtain a cloud-free clear image after the thin-cloud correction, comprising: calculating the disturbance increment of the thin cloud radiation of each wave band according to the thin cloud independent components and the corresponding cross-wave band coefficient vector; And subtracting the disturbance increment of the thin cloud radiation from the corresponding wave band of the thin cloud image to be corrected pixel by pixel to obtain a cloud-free clear image after the thin cloud correction.
8. A visible light remote sensing image thin cloud correcting device is characterized by comprising: The data acquisition module is used for acquiring a thin cloud image to be corrected and a cloud-free reference image, and preprocessing the two images; The independent component analysis module is used for constructing a differential domain based on the preprocessed thin cloud image to be corrected and the cloud-free reference image, and carrying out independent component analysis on differential domain data to obtain a plurality of independent components and a mixing coefficient matrix; The independent component judging module is used for calculating the low-frequency energy duty ratio of each independent component, identifying the thin cloud independent component according to the low-frequency energy duty ratio, and extracting a cross-band coefficient vector corresponding to the thin cloud independent component from the mixed coefficient matrix; And the thin cloud correction output module is used for carrying out thin cloud stripping processing on the thin cloud image to be corrected based on the cross-band coefficient vector to obtain a cloud-free clear image after thin cloud correction.
9. An electronic device, comprising: A memory for storing a computer program; a processor for executing the computer program; The method for correcting the thin cloud of the visible light remote sensing image according to any one of claims 1 to 7 is realized when the processor executes the computer program.
10. A computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and the computer program when executed implements the visible light remote sensing image thin cloud correction method according to any one of claims 1 to 7.

Description

Visible light remote sensing image thin cloud correction method, device, equipment and medium Technical Field The invention relates to the technical field of image processing, in particular to a method, a device, equipment and a medium for correcting visible light remote sensing image thin cloud. Background The visible light remote sensing image is widely applied to scenes such as agricultural monitoring, urban fine drawing, change detection, target recognition and the like, is obviously influenced by cloud layers, and is smooth in space and free of obvious boundaries at present, so that cloud coverage proportion in the global scope is long-term at a higher level, cloud removal becomes high-frequency requirement in remote sensing pretreatment, and compared with the situation that surface information is lost due to thick cloud shielding, thin clouds (including mist, thin curly clouds and the like) have semitransparent characteristics, and the visible light is usually expressed as radiation disturbance of high brightness, contrast reduction, color shift and texture blurring caused by scattering of atmospheric suspended particles, so that quantitative correction of the thin clouds is more challenging. At present, a plurality of technical routes are formed for thin cloud correction of visible light remote sensing images, but the problems of stable separation of the thin cloud radiation disturbance, cross-source and cross-resolution adaptability, interpretation and controllability of results and the like still exist, and reliable and consistent correction results are difficult to obtain under complex scenes and various data conditions. Disclosure of Invention The embodiment of the invention provides a thin cloud correction method for a visible light remote sensing image, which effectively solves the problems of image blurring and surface information distortion caused by thin cloud, remarkably improves the definition and usability of the thin cloud image, and improves the quality and application value of the remote sensing image. In a first aspect, an embodiment of the present invention provides a method for correcting a thin cloud of a visible light remote sensing image, including: acquiring a thin cloud image to be corrected and a cloud-free reference image, and preprocessing the two images; Constructing a differential domain based on the preprocessed thin cloud image to be corrected and the cloud-free reference image, and performing independent component analysis on differential domain data to obtain a plurality of independent components and a mixing coefficient matrix; calculating the low-frequency energy duty ratio of each independent component, identifying the thin cloud independent component according to the low-frequency energy duty ratio, and extracting a cross-band coefficient vector corresponding to the thin cloud independent component from the mixed coefficient matrix; and carrying out thin cloud stripping treatment on the thin cloud image to be corrected based on the cross-band coefficient vector to obtain a cloud-free clear image after thin cloud correction. Further, the preprocessing the two images includes: Resampling the cloud-free reference image to enable the spatial resolution of the resampled cloud-free reference image to be consistent with that of the thin cloud image to be corrected; and carrying out band-by-band radiation unification treatment on the resampled cloudless reference image so as to ensure that the band of the uniformed cloudless reference image is consistent with that of the thin cloud image to be corrected. Further, the band-by-band radiation uniformization processing of the resampled cloud-free reference image includes: According to the thin cloud image to be corrected and the resampled cloud-free reference image, calculating the total difference of each pixel in all wave bands; Selecting a stable pixel set based on the total difference, and calculating a radiation consistency parameter in a stable pixel area by adopting a least square method; And carrying out band-by-band correction on the resampled cloud-free reference image by using the radiation uniformization parameters to obtain the cloud-free reference image after radiation uniformization. Further, the constructing a differential domain based on the preprocessed thin cloud image to be corrected and the cloud-free reference image, and performing independent component analysis on differential domain data to obtain a plurality of independent components and a mixing coefficient matrix, includes: Performing band-by-band difference on the preprocessed thin cloud image to be corrected and the cloud-free reference image to obtain a multiband difference image; Constructing an observation matrix of the multiband differential image, wherein each column of the observation matrix corresponds to a multiband differential vector of one pixel; and carrying out independent component analysis on the observation matrix to obtain three in