Research On The Algorithm Of Ground Object Information Recovery For Thin Cloud Contaminated Remote Sensing Images Based On Image Transformation And Transfer Learning

The remote sensing images taken by satellite imaging devices are often covered by clouds.Obtaining ground object information of remote sensing image under clouds via information recovery technology can effectively enhance the clarity of the remote sensing images and improve the interpretability of the cloud cover area.Based on image transformation and transfer learning,This paper researches the algorithms of ground object information recovery based on image transformation and transfer learning by using the thin cloud contaminated remote sensing images obtained by Landsat and HJ-1 satellites.The main research contents and innovations are as follows:1.The basic parameters and the main functions of Landsat series and the HJ1 A/B satellite are introduced.Then the purpose and significance of recovering ground object information of remote sensing image under thin clouds area are elaborated.Subsequently the present situation of recovering ground object information at home and abroad are analyzed.2.In order to facilitate the further processing of remote sensing images,some related theoretical knowledge of remote sensing image pre-processing is introduced including remote sensing image atmospheric correction,geometric correction and image registration.3.A ground object information recovery algorithm for remote sensing images covered by thin cloud is presented based on the modified NSCT and T-SVR..Firstly,the improved Nonsubsampled Contourlet transform(NSCT)is used to decompose the multi-source and multi-temporal remote sensing images.Then change detection is implemented on the high frequency part of the decomposed image.Based on the low frequency coefficients of the other bands in the target image,the Transfer support vector regression(T-SVR)is used to predict the low frequency coefficients in the change region.For no change in the region,the low frequency coefficients is predicted by using T-SVR according to the low frequency coefficients of cloudless reference image.Finally,the reconstructed image is obtained.The experimental results show that the proposed algorithm makes full use of the effective information of the remote sensing images,and the accuracy of the information retrieval is high.4.A ground object information recovery algorithm for thin cloud contaminated remote sensing images is presented by combining M-NDTCWT and T-LSSVR.Firstly,multi-resolution decomposition of multi-source and multi-temporal remote sensing images is performed by using multi-directional nonsubsampled dual-tree complex wavelet transform.The decomposed high frequency coefficients of the ground objects of the thin cloud images are primarily classified by using Bayesian method.Then the transfer least square support vector regression model is trained to obtain the model parameters by using the domain adaptive learning of the low frequency coefficients of each class of ground objects.Finally,the low frequency coefficients of the thin cloud-contaminated images are predicted by using those of the cloudless reference images.The thin clouds are removed and the ground object information of the thin cloud contaminated images is recovered.Experimental results show that the ground objects recovered by the proposed algorithm have clear spatial details and small spectral distortion.Especially for the thin cloud contaminated remote sensing images with seasonal variation of ground objects,the proposed algorithm can effectively recover the ground object information contaminated by thin clouds.