Research On Cloud Removal Technology Of High Resolution Remote Sensing Image

Nowadays,high-resolution remote sensing images have been widely applied to various fields.Due to the influence of external factors such as light and the atmosphere,the image clarity of the remote sensing system may be reduced.For example,cloud pollution in remote sensing images will affect the interpretation of terrain information.If it is a thin cloud noise,there is no need auxiliary image for noise reduction.For thick cloud noise,the feature information is almost completely blocked,and the cloud is removed by the auxiliary image.The thin cloud and thick cloud removal technology proposed in this thesis for high-resolution visible light remote sensing images has good effect and practicality.The specific research mainly includes the following two parts:For the removal of thin cloud noise in visible light images,this thesis proposes an improved cloud removal algorithm based on multiscale wavelet analysis.This algorithm can improve the inaccuracy of thin cloud information discrimination by traditional wavelet method.First use the dark channel prior algorithm to process the dark channel of the visible light image to obtain the dark channel image,and then perform multi-scale wavelet decomposition of the visible light image.After decomposition,the low-level detail coefficients are reconstructed to obtain high-frequency components.High-frequency components correspond to the feature detail information of the image.The dark channel image is weighted and combined with the high-frequency components to enhance the feature information.The high-level detail coefficients are reconstructed to obtain low-frequency components.The high-frequency emphasis filter is used to suppress low-frequency cloud noise,and the approximate coefficients are reconstructed.Finally,the processed high-level detail coefficient reconstruction map,low-level detail coefficient reconstruction map,and approximate coefficient reconstruction map are combined to obtain cloud removal image.For the different locations of the thick clouds in the two remote sensing images at different phases,this thesis proposes an image fusion algorithm based on wavelet analysis.This algorithm solves the problem of cloud residues in traditional wavelet fusion de-clouding method.Due to the large spatial resolution of the data set and the obvious feature information,the fusion method in this thesis improvements have been made.The main improvement is the fusion rule for low-frequency components.This thesis proposes to apply a dark channel-based cloud detection algorithm to the fusion rules for low-frequency components.First perform dark channel processing on two images with different phases to obtain their respective dark channel images,and then perform wavelet decomposition on the two images,reconstruct the respective low frequency components,and fuse the low frequency components of the two images by judging the position of the cloud.For high-frequency components,local energy fusion algorithms are used for fusion,and reconstruction is performed after fusion.Finally,the high-frequency component and the low-frequency component after fusion are superimposed to obtain a cloud removal image.Through the analysis of the experimental results,it is verified that the proposed algorithm can effectively remove the thin cloud noise and thick cloud noise of the remote sensing image.In this thesis,several methods of thin cloud removal are compared,and the results show that the algorithm proposed in this thesis has achieved good results in visual analysis and objective evaluation,and improves the inaccuracy of thin cloud information discrimination by traditional wavelet method.The similarity between the removed image and the original image is higher,and the visibility is better,which effectively solves the problem of cloud residue.