On-orbit Geometric Calibration And Bundle Block Adjustment Of Spaceborne Three-line Scanner

Remote sensing and earth observation technology is the one of the most significant method for mankind to collect the information of the earth surface. More and more remote sensing satellites are launched into orbits as tremendous progress have been made in the space remote sensing. The resolution and geolocation accuracy have been largely improved. WorldView-3, developed by Digital Global and launched at August13th2014, is the third satellite of WorldView series. It has the highest resolution of0.33m among the commercial satellites all over the world. The geolocation accuracy, which is better than3.5m, reaches the state-of-art level. China had also started their high resolution earth observation satellite project at May2010. Remote sensing and earth observation technology became one of the research hotspots in the remote sensing field. Geolocation of remote sensing satellite is significant since high accuracy of geolocation is prerequisite condition before wide applications.Orbit and attitude determination, satellite time and postprocess procedure include sensor on-orbit calibration and bundle block adjustment are the main factors which affect the geolocation accuray. Restrained by the hardware and orbit and attitude determination technology, the accuracy of most remote sensing satellites in China can not meet the accuracy requirement of wide applications, thus the postprocess procedure such as sensor on-orbit calibration and bundle block adjustment became the most important and effective way to improve the geolocation accuracy of these satellites.Methods of on orbit calibration of spaceborne linear array scanner and bundle block adjustment are firstly intrduced in this paper, then four research subjects are proposed:(1) On-orbit calibration of sensors with multi-strip imagery based on mass control points with low accuraciesIn this subject, we studied the imaging geometry and orientation model of linear array sensors, introduced the relative on-orbit calibration parameters according to the sensor imaging pricinple, and investigated the correlations among those parameters and effect on geolocation accuracies. Strategies of eliminating or decreasing those correlations were proposed. Given that ground control points are hard to extracted and the on-orbit calibration is highly dependent on the ground test field, an automatic control points extraction technology is applied in combined on-orbit calibration with multistrip imagery. Plenty of real data collected by three-line camera of ZY-3are used to test the proposed calibration method. Prelimilary result have shown that this new calibration method does not need any test field or high accuracy GCPs. Only medium accuracy ortho images and DEM are needed Thus, this method can substantially decrease the high cost of test field construction and maintenance; and a GCP automatic matching system instead of the traditional artificial identification method can make the entire calibration process fully automatic, which could significantly improve its efficiency.(2) Bundle block adjustment with auto-calibration using linear array imagery base on mass control points with low accuraciesThe basic pricinple and theory of bundle block adjustment with auto-calibration are given. The calibration parameters are introduced and the calibration model are established(4classes of compensations include a total of21parameters model which proposed by Professor Brown in America and total12pure mathematic parameters model which proposed by professor Ebner in German). Parameters with no correlations or have actual geometry and physical meaning should be chosed for auto-calibration to compensate the systematic error more effectively. Different combinations of IOPs in the self-calibration bundle adjustment with one strip of CBERS-02B real data were tested in this paper. Tests results show that the accuracy of the bundle adjustment with IOPs was improved by two to three pixels when compared to the adjustment without IOPs.(3) Bundle block adjustment of spaceborne linear array imagery with ground control pointsThe classic blunder detection and identification method and weight selection strategies are introduced in this subject. Proper orientation model and blunder etection and identification strategies are applied to perform bundle block adjustment experiments with three-line array imagery of TH-1and ZY-3satellites. Test results demonstrated that higher accuracy can be obtained by using orientation image model than other models while dealing with long orbit data. As to ZY-3imagery, Accuracy of bundle block adjustment with systematic error compensation model using mass control points with low accuracies can achieve the accuracy of better than15m in plane and12m in height.(4) Bundle block adjustment of multistrip spaceborne linear array imagery without ground control pointsThe algorithm pricinple of bundle block adjustment and the weight determination strategies are discussed in this subject. As to multistrip imagery, overlap area between adjacent strips can be used as constrains in the whole bundle block adjustment, which can improve the stabilities of bundle adjustment without ground control points. The storage, management and calculations of massive data and inversion of super large matrix cause by multistrip dataset are also discussed. A total of13strips of ZY-3TLC imagery were used to perform bundle block adjustment without ground control points.The preliminary test results show that multi-strips bundle block adjustment without GCPs can achieve accuracies of about8m in planimetry and5m in height, and the nationwide bundle block adjustment is considered practical and feasible.