Research On Removing Thin Cloud In A High-resolution Remote Sensing Image

High-resolution remote sensing images provide valuable data source for investigation and monitor on Land,forestry,agriculture and environment.But in practice,high-resolution remote sensing images will inevitably affected by cloud cover,so it is very important how to interpret information of remote sensing data accurately and make full use of remote sensing data.Based on practical applications,the several common methods of cloud-removing are deeply discussed and systematically researched.Combined with tasseled cap transform(TC transform),a improved version of haze-optimized transformation(HOT)as a new method of cloud-removing is proposed and used to obtain the region of cloudless sky in the image automatically.Due to the disadvantages of traditional HOT method,low confidence on non-vegetation region,the vegetation region with high confidence coefficient should been extracted on normalized differential vegetation index(NDVI).Therefore,the vegetation region can be used to invert the haze noise information of the whole image and achieve the aim of cloud-removing.Finally,when the underlying surface is water,the algorithm of cloud-removing is automatically implemented through ATCOR software.Based on ENVI/IDL second development interface,a plug-in program of cloud-removing automatically is developed to remove haze influence operationally.The high resolution remote sensing images of Beijing and Henan are experimental subjects,which are filmed by GF-1 and GF-2 sensors.There is a mess of haze in these two images,however,the final images are turned to be perfect by using the new cloud-removing method.Then the results in subjective sight check and objective statistics methods are analyzed.The experiment proved that the haze region of the original can be effectively detected with the method that combines HOT and TC transform with NDVI.And it can figure out the problem of low degree of automation and colorful building distortion.The result of improved algorithm is better than the result of non-improved algorithm.Based on thin cloud removal on images,more studies of the atmospheric correction are made in this thesis.Vector 6S codes are utilized to calculate scattering and absorption effects in the radiation transfer process respectively.The former is stored in the Look-up tables(LUTs),containing variety of combinations from the atmospheric models,aerosol types,imaging geometry,et al.The latter is expressed as the experience formulas,of atmospheric transmittances,calculated in the conditions of six kinds of atmospheric models.The size of look-up tables are much reduced because of the separated treatment of scattering and absorption.Atmospheric correction based on LUTs for reflectance can reflect the real information on the surface of the earth.Atmosphere correction LUTS based on 6SV model are used for atmospheric correction on GF2 image after removing thin cloud,and the results conform to the spectral curve features.It illustrates that cloud removal algorithm can remove thin cloud and keep the original image feature information,and the LUTs is simple,effective and quick to implement the atmospheric correction for GF2 images.