Theory And Method Study Of Geopositioning For High-resolution Satellite Imageries Based On Lie Algebra

With continuous development of space remote sensing technologies,high-resolution earth observation systems are gradually transiting from special-purpose large platforms to a generalpurpose small platforms,from single satellite observation to multi-satellite network fusion detection,from ground professional processing to on-board real-time processing.The application of high-resolution satellite remote sensing image has put forward new requirements for the universality,timeliness and high precision of data processing.The thesis focuses on the new theory and method of geometric positioning of high-resolution satellite remote sensing image.The theory and method of high-resolution satellite remote sensing image positioning based on Lie algebra are mainly studied,including the modeling of external orientation elements of attitude,the construction of rigorous imaging model,the adjustment of stereo positioning and regional network,and the generation of rational function model and etc.The research results theoretically deepen and expand the current satellite remote sensing image positioning theory and method,and improve the uncontrolled positioning accuracy through the innovation of theory and algorithm in application.The main works are addressed as follows:1.Firstly,the development status of high precision positioning theory and technology of highresolution earth observation system has been deeply analyzed.In view of the problems and shortcomings in data processing when Euler angle and quaternion are traditionally used to represent the attitude of linear array satellite images,the Lie group/Lie algebra is applied to the attitude representation of high-resolution satellite remote sensing images,and the corresponding geometric positioning theory system and new data processing method are established.2.The basic algorithm models and numerical calculation methods of high-resolution satellite remote sensing image based on Lie algebra representation and modeling,collinear condition equation linearization,linear interpolation exterior orientation element modeling and linear interpolation collinear condition equation linearization are proposed.The theory and method system of linear array satellite image processing based on Lie algebra are established.The results show that compared with Euler angle and quaternion,Lie algebra interpolation is more accurate and the interpolation result is complete and smooth;the accuracy of resection calculation is consistent,and the efficiency is significantly improved.It is feasible to use lie algebra to represent the attitude elements of aerial antenna array photogrammetry.3.The rigorous imaging model of high resolution satellite remote sensing image based on Lie algebra attitude representation and the interpolation method of arbitrary scanning line image of linear array satellite are proposed.Under the condition of Lie algebra attitude representation,the imaging geometric relationship between ground target and satellite image is established.The theory and method of single image location,stereo image location and multi strip image location based on Lie algebra are proposed.The accuracy of Lie algebra stereo positioning is verified by using three kinds of terrain data of TH-1 satellite located in plains,hills and mountains.It is proved that compared with the traditional Euler angle,Lie algebra stereo positioning has higher accuracy,better stability and stronger practicability.4.An block bundle adjustment model and method without no ground points for high resolution satellite images based on Lie algebra is proposed.The EFP and orbit piecewise polynomial fitting method based on Lie algebra are established,and the specific calculation process is given.Compared with the traditional Euler angle representation,it is verified by the actual data of TH-1 01 and 03 located in Southwest China that the block bundle adjustment based on Lie algebra is more sensitive in the detection of system error,and the positioning accuracy can be improved under two different adjustment models,especially the efficiency improvement.5.A method to solve RPC parameters based on HEIV model is proposed.And the method is aimed at the problem of unequal precision noise in the design matrix elements of RPC parameter estimation and the systematic deviation in the conventional least square adjustment method.Taking the minimum Mahalanobis distance as the adjustment criterion,the total least squares method considering the error characteristics of all elements of the design matrix is used instead of the traditional least square method,the relevant calculation formulas and numerical calculation methods are derived.The actual data validation of TH-1 satellite in three typical terrain of plain,hilly and mountainous area shows that this method has obvious advantages in image correction accuracy compared with the existing direct least square method,L-curve ridge estimation least square method and truncated SVD estimation method,and can significantly improve the estimation accuracy of RPC parameters.6.Based on the satellite image data and RPC parameters,a method is proposed to reverse the parameters of the precise imaging model,such as azimuth elements,orbit,attitude,etc.In the method,the satellite image and RPC parameters can be used to calculate the internal and external azimuth elements of the corresponding imaging time of the satellite image area without the initial value of the mission.The rational function model can be used to reverse the rigorous imaging model,and the key link of the mutual transformation between the rigorous imaging model and the rational function model is opened up.Through data reverse experiment by several scenes of TH-1 01 satellite,the result is that the maximum residual error of exterior azimuth line element is 0.96 m,the maximum residual error of angle element is 0.95 angular second,and the maximum residual error of internal azimuth is 0.42 pixel.The rationality of the theory and the practicability of the method are verified.It provides a new way to realize long strip adjustment of rational function model,high precision positioning with less control / no control and multi-sensor joint processing.