The Research Of Relative Radiometric Correction On ZY-3 Satellite Image Based On Landsat8 Image

At present,a series of achievements have been made in photogrammetry field.But the phenomenon of color distortion often appear in the process of DOM(Digital Ortho Model)production.To deal with the color and tonal inconsistency problem,a method of digital image processing was proposed traditionally.Which was discarded in the end,as it can cause the loss of image details.Therefore,this article proposed a scheme to improve the radian quality of domestic high resolution satellite image.From the perspective of distortion source,it will introduce an open base image,which is of high quality of radiation.The main research is as follows:(1)Making radiation calibration base image.Execute the model of Top Of Atmosphere(TOA)and Semi-empirical kernel function BRDF calibration for,andsat8 images.The former is used to eliminate or decrease the atmospheric influence on imaging quality of radiation.The latter is to reduce the effects on radiation caused by sun angle and angle of observation.In the BRDF calibration system,radiation control points are added in,and the samples are classified according to the solar incident angle,which takes the terrainfactors into account,thus improving the stability of solution.(2)Relative radiation calibration for the domestic high resolution satellite images.A simple relative radiation calibration method,Invariable Targets Method is put into practice for ZY-3 images as the base image is made.This paper expounds the difficulties of selecting the invariable targets,and the introduce a method of automatic selecting vegetation and bare ground.According to the relationship between different phase of images,it is feasible to calibrate the radian quality of ZY-3 images to meet the base image level.(3)Experiments for verifying the rationality and feasibility of scheme.Three indicators are put forward to evaluate the scheme,the correlation coefficient(r),describing the relationship between the source image and base image,the information entropy(E),evaluating the valuable information in images,the principle signal angle(PSA),measuring the similarity of source and base images.The PSA,which can ensure the orthogonal property,the maximum of SNR as well,is proposed on the foundation of the MNF(Minimum Noise Fraction)model.Taking a scene of ZY-3 image for example,whose serial number is 032-110,the improvement for information entropy are corresponding 13.34%?15.57%?8.57%for blue?green and red bands.The principal signal angle is near to 0 between the ZY-3 image and Landsat8 image.In addition,in terms of visual effects,the quality is greatly enhanced.The images share consistent color,with overlap zone.Color is very close to the real object features.It indicates that the research of this paper is a feasible scheme for the production of domestic high resolution satellite calibration.