Precise Geometric Processing Of High Resolution Optical Satellite Imagery

High resolution optical imaging is a significant trend of global earth observation technology development, and also the main direction of remote sensing industry in our country. Both internal and external geometric quality of high resolution optical satellite imagery are essential for data deep processing and multiple applications. This paper summarizes recent advances in the field of satellite image geometric processing at home and abroad, on the basis of whole process imaging systematic errors analysis, proposes a set of key technologies of image high precision geometric processing.This paper provides detailed illustration of model algorithm design and solving procedures of these key technologies. First, a relaxation method for on orbit geometrical calibration is proposed, which is suitble for triangular mechanical staggered stitching imaging sensor. Improved conjugate gradient block adjustment is used to solve the internal and external calibration model parameters respectively; next, the detection and compensation of attitude high-frequency jitter of satellite platform are considered in the geometric calibration and the direct positioning of calibrated image. The attitude and positioning accuracy are optimized by time-division imaging data or high-frequency angular displacement data; finally, on the basis of geometric calibration and jitter compensation, the image data is geometric rectified by rigorous imaging model with the large tilt angle imaging and terrain variation positioning compensation taken into consideration, and the high precision geometric rectification by rigorous transformation model based on line features and generalized point strategy.Specifically, the research work in this paper mainly include the following 5 aspects:(1) A high precision on-orbit geometric calibration method based on multi-scene image and multi-chip TDICCD triangular mechanical staggered stitching imaging sensor is proposed in this paper. In order to ensure the accuracy of geometric calibration while reducing the correlation between parameters, the geometric calibration model is optimized, and multi-chip calibration method based on pointing angle model is used for the internal calibration. Based on the polynomial point-by-point fitting of satellite attitude and orbit data with weight estimation, the overlapping connection relationship between multi-scene images and multi-chip TDICCDs are fully utilized for construct-ing the geometric internal and external calibration error equations with constraints, and they are iteratively solved through relaxation method block adjustment. This geometric calibration method can effectively and stably eliminate various system errors of the imaging system, and reduce the image internal geometric distortion and improve the absolute positioning accuracy.(2) A conjugate gradient block adjustment method with preprocessing of improved incomplete Cholesky factorization is proposed, which is applied for the adjustment calculation of high precision on-orbit geometric calibration. The conjugate gradient block adjustment method is deduced by the variation method in theory, and the design of preprocessing matrix is optimized. In each iteration of the relaxation method geometric calibration, the detailed solution process is proposed for internal and external calibration by block adjustment based on improved conjugate gradient method. This method can optimize the structure of the coefficient matrix of the normal equation, improve the calculation efficiency of geometric calibration and has good stability, under the premise of ensuring the adjustment precision.(3) A method of high frequency jitter detection of satellite platform and image direct positioning compensation based on time-division imaging imagery and high-frequency attitude measurement data of angular displacement equipment is proposed. Ultilizing the multispectral imagery from multi-spectral combination imaging and panchromatic imagery from triangular mechanical staggered stitching imaging, respect-tively, the registration error curve and the CCD misalignment curve are generated by dense matching. Then, the high frequency jitter model is constructed by multi-sine function fitting, and it is applied for the direct positioning compensation of rigorous imaging geometric model based on probe pointing angle; With the angular displacement high-frequency attitude measuring device, the preprocessing and jitter detection based on the high-frequency angular increment data are researched. The phase and spectrum jitter law of the satellite platform in pitch, roll and yaw directions are analyzed, and applied for the direct positioning compensation of rigorous imaging geometric model.(4) A high precision rigorous geometric rectification method for high resolution optical satellite image is proposed, which takes into account the direct positioning compensation under large tilt angle imaging mode.To solve the problem of geometrical accuracy degradation of large tilt angle imaging data, this paper systematically analyzes the geometric characteristics of large tilt angle imaging, and quantitatively studied the variation of the ground resolution and direct positioning accuracy caused by the large inclination angle imaging, the attitude angle error and the terrain variation. On the basis of inverse solution method of rigorous imaging geometric model, rigorous geometric rectification model with image point error compensation of large tilt angle is constructed with control data, and the resolution of the rectified image is normalized, this method can effectively eliminate the loss of positioning accuracy caused by tilt imaging, thus improve the internal and external geometric quality of imagery.(5) A high precision geometric rectification method of high resolution optical satellite image is proposed based on rigorous model and by line feature vector control and generalized point strategy. First, the rigorous transformation model based on line features is constructed based on generalized point strategy. Then, the preprocessing of model parameters is performed by the line-based six-parameter transformation model, and their precise values are solved through least squares overall adjustment. Finally, the high-precision geometric rectification process by line-based rigorous transformation model is provided, which can effectively take into account the impact of satellite imaging tilt angle and topography. In addition, this paper also analyzes the influence of control line’s types, quantity, horizontal and vertical distribution on model accuracy. They can provide beneficial technical reference for reasonable selection of control lines to further improve the precision of geometric rectification.Experimental comparison and analysis are made using imageries captured from Chinese TH01 and CRSS satellites imageries, as well as American IKONOS and GeoEye-1 satellites imageries, which demonstrate that, the methods proposed in this paper can not only fully tap the positioning potentials of imaging system, but also improve the geometric quality of imageries under multi-imaging modes with good stability and computational efficiency. They can provide effective technology support for the construction of satellite data processing system and multilevel applications.