Research On Sparse Clouds Removal Method By RPCA In Remote Sensing Image

Remote sensing imaging technology has become a key research direction for many countries and research institutions in recent years.However,due to the technical limitations,most remote sensing images cannot avoid being covered by clouds.Therefore,it is necessary to adopt some methods to remove the cloud within a local area.The traditional cloud removal methods always deal with the thick cloud and thin cloud separately.But thin clouds often appear with thick clouds.It is difficult to remove the cloud by a single method.For this reason,an effective method to remove clouds in a small distribution range is studied.Due to the area of this type clouds is smaller than the large-scale remote sensing images,the clouds has significant sparse characteristics.So,an improved robust principal component analysis(RPCA)algorithm of the low rank matrix recovery(LRMR)algorithms that can be applied to the cloud removal of remote sensing image is proposed based on the sparsity characteristics of clouds.(1)The cloud removal theory of remote sensing image and image registration are studied.The basic principles of cloud removal and some common cloud removal methods are described.In addition,aiming at the registration problem in remote sensing image preprocessing,an improved remote sensing image registration technology based on SIFT feature is proposed.The attribute reduction of rough set with CQPSO is introduced to reduced dimension of feature vectors.The experimental results show that improved SIFT algorithm can effectively improve the accuracy of registration.(2)The RPCA common algorithms are analyzed and improved.Common algorithms of RPCA are studied.In order to solve the problem that the accuracy and speed slows down when dealing with large scale complex matrices,an improved RPCA algorithm is proposed.A new RPCA model is constructed based on approximate L0 norm,and a multilevel low rank approximation technique based on coarse SVD is adopted.It is demonstrated by experimental result,the improved RPCA algorithm has better performance in both operational efficiency and algorithm accuracy compared with common algorithms.(3)Application on cloud removal based on improved RPCA is studied.In order to solve the problem that the common cloud removal methods can't remove the complex clouds,a method to remove clouds based on improved RPCA is presented.First,a low rank observation matrix is constructed.Then,an improved RPCA algorithm is used to remove sparse clouds.Finally,a cloudless remote sensing image sequence is reconstructed.It is demonstrated by experimental result that the cloud removal algorithm of remote sensing image based on improved RPCA can remove the cloud in complex sparse cloud occlusion.It has better performance in subjective visual and objective indicator.