High Resolution Satellite Image Fine Processing Method For 3D Scene Of Ore Concetration Area In Qinghai-Tibet Plateau

Tibetan Plateau is the important mineral resources area of china.Nevertheless,it is quite difficult to prospect and explore mineral resources in that area,due to the bitter natural environment and insufficient geographic information.With the rapid development of space remote sensing and computer technology,the high-resolution(HR)remote sensing satellite system plays an efficient role in mineral resources exploration.Using HR remote sensing image to model 3D field scene has great advantages on the speed of data collection,field workload and harsh natural environment such as Tibetan plateau.Considering the 3d scene characteristics of the Qinghai-Tibet Plateau ore concentration area,and the special problems involving the 0.5-meter HR satellite remote sensing image processing,this thesis made in-depth investigation to probe into building up imaging geometric models for Geoeye-1 and WorldView-2 Imagery,3D Scene Re-construction,imagery Ground Control Point(GCP)deploying method,3D Scene Re-construction and information integration of the Metallogenic District,and other key issues.The research efforts and innovations are listed as follows;1.Analyzing the Push-broom Linear Array CCD imagery principle,the author built up satellite rigorous sensor model(RSM),which considers the velocity aberration,the optical path delay and the atmospheric refraction,and create a new physical inverse model based on a double-iterative method.With these models to conduct uncontrolled single image positioning,Geoeye-1 single image can have direct positioning accuracy of 5.4m(CE90)and stereo 3.2m;WorldView-2 single image plane accuracy is 6.7m(CE90)and stereo 5.0m.The inversion model takes into account the convergence conditions of the image object space and the image space,which is more accurate than the existing single iterative model and can meet the requirement of 0.5 meter high resolution remote sensing satellite image back projection.WorldView-2 image experiment results show that,the mean square error of the image space is 6.7 pixels,in the object space is 3.7 m,and the correctness of the algorithm is verified by the standard positioning accuracy of the WorldView-2 satellite.2.Based on the 3D re-construction principles for Rational Function Models(RFMs),the author proposed a direct RPC affine transformation correction model,and used the model to optimize images for WorldView-2,reducing the row error from 3.6 Pixels to 1.54 Pixels,and column error from 4.30 Pixels to 1.81 Pixels,and improving imagery 3D intersection plane precision from 4.0m to 2.3m,and vertical accuracy from 3.8m to 1.8m.It is shown that this model is valid and effective for HR remote sensing images.The corrective accuracy is consistent with the RPC indirect correction model.The optimized RPC coefficients can be used directly in commercial softwares,and the indirect correction model has better applicability.The covariance analysis of the design matrix column vector in the process of solving the RPC coefficients is carried out by using the correlation coefficient method.It is verified that the RPC parameters can be optimized by the correlation coefficient method even in the terrain-independent scheme calculation,and reduce the subsequent calculation.3.Given that there are no high national level sufficient vertical control points,but those with uneven distribution around the mountainous Qinghai-Tibet Plateau,which leads to considerable errors of conventional GPS fitting,the author made use of EGM2008 gravity field models to convert baseline geodetic height difference into normal height difference,and had rigorous adjustment in the GPS vertical control network to solve for elevation of unknown points.The calculated results were better than conventional GPS elevation fitting.4.The PRC block adjustment experiment in terms of different control points schemes was made on different software platforms,and Geoeye-1 and WorldView-2 imagery 3D re-construction precision showed that Rational Function Models(RFMs)for mountainous plateau areas still featured high precision,in line with rigorous geometric models.For multiple images,image pairs and/or strips,single control points would not be able to control the 3D re-construction precision around the whole area,and control points deployed at the four corners around the whole image and in the center made the optimal scheme.These points connected formed up an integrated whole,with its RPC block adjustment to obtain stable optimal solutions.Under no-control conditions,Geoeye-1 satellite 3D positioning plane precision was up to 3.0m,with vertical accuracy up to 6.0m while WorldView-2 plane accuracy reached lm,with vertical accuracy up to 3.0m.Under controlled conditions,Geoeye-1 satellite and WorldView-2 satellite 3D positioning boasted the optimally accurate plane and elevation within lm.The experiment indicated that both of the two satellites meet the requirements for imagery with plane and elevation precisions of 1:10,000 and 1:5,000 respectively.The experiment findings verified that Geoeye-1 and WorldView-2 3D imagery systems were able to do 3D scene re-construction for mountainous Qinghai-Tibet Plateau with high altitude and dramatically fluctuating topography,in the same high accuracy as they do for the plain areas.5.Investigating some HR satellite imagery space information extraction methods,the author had detailed investigation of the DEM,DOM and DLG making processes.For making DOM,an integrated index method was made to have comprehensive evaluation of image fusion effects,so as to objectively ascertain the fusion effects of different integration algorithms,and to provide some new idea for image fusion quality evaluation.For making DEM,image archives for the same coverage in different periods were employed to form cross-track 3D images,whose DEM extraction precision is able to meet the high altitude/elevation precision requirements for high land areas in the 1:50,000 mapping specifications.6.With DEM,DOM and DLG data extracted from high resolution images,the author explored the 3D scene modeling methods under the SpaceEyes 3D platform,and employed some plug-in technology for mines to develop a set of 3D mine information integration experiment system,providing an important foundation for digitalization construction of mines in the plateau area.