Study On Theory And Method Of Geopositioning For High-Resolution Satellite Imagery Based On Quaternion

High precision positioning is the basis for remote sensing image processing and geospatialinformation geometry obtaining, is the support of topographic mapping use of high resolutionsatellite imagery. In order to solve the problems which existed in the geometry processing ofhigh resolution satellite imagery, this dissertation, which uses quaternion as a mathematical toolto describing the imaging attitude, makes a comprehensive study on the theory and methods ofgeopositioning for high resolution satellite imagery based on quaternion, which includes themodelling of exterior orientation parameters (EOPs) using quaternion, the establishling of therigorous geometric model, quaternion bundle block adjustment based on multi sensors, etc. Themain achievements and innovations are summarized as following:1. Based on the advantages of three dimensional rotation described by quaternion and theimaging characteristics of the linear array pushbroom image, unit quaternion is used to describethe attitude of the image, and two kinds of exterior orientation parameters (EOPs) model areestablished based on quaternion differential equation (QDE) and quaternion spherical linearinterpolation (QSLERP) respectively.2. A qualitative and quantitative analysis method of the precision and stability for EOPssolution using SNR is proposed. Simulation image and real high resolution satellite image areused to analyze and test the effect of positioning accuracy for the correlation of EOPs, theresults and conclusions of these experiments can be use for the reference of the research on thegeopositioning for high resolution satellite imagery based on quaternion.3. The rigorous geometric model based on quaternion is established. In order to overcomethe obvious correlation when quaternion is used to describe the attitude of the image, a methodof using three independent parameters which deduced from quaternion to solve EOPs isproposed, and the corresponding error equations are derived. So that, the indirect adjustmentwith constraints is not required, the existed adjustment mehod in photogrammetry can be usedto solve EOPs, and it is convenient for integrating into the existed photogrammetry software.4. To further overcome the correlation of EOPs, Tikhonov regularization method isimproved. The prior information of correlation, which is reflected in signal to noise ratio, is useto construct regularization matrix and parameters in this improved method .5. Quaternion integrated sensor orientation using ephemeris and attitude data is researched,and the time synchronization error between the ephemeris and attitude data and the image ispreliminary discussed. A method of raising positioning precision without ground control points(GCPs) through corrected the image corrdinates in flight direction is presented, and detailed experiments with SPOT5 HRS images are used to validate the validity and rationality of thequaternion integrated sensor orientation and error corrected methods. The results show that thecorrected the image corrdinates in flight direction can improve the integrated sensor orientationaccuracy evidently.6. Considering the independent parameters, a solution of exterior orientation parametersfor high resolution satellite imagery based on quaternion differential equation is proposed.Experimental results indicate that this solution, which can avoid computation of trigonometricfunctions completely, is right, stable and adaptive, and the orientation precision can beimproved when these three independent parameters are used.7. A solution of EOPs for high resolution satellite imagery based on quaternion sphericallinear interpolation is improved on the base of independent parameters. Robust solution ofEOPs using iteration method with variable weights is used when there are some coordinateblunders in GCPs, and quaternion stereo orientation method is experimented to verify theeffectiveness and adaptability of these methods.8. The quaternion bundle block adjustment based on multi sensors is studied when theephemeris and attitude datas are regarded as observed value with weight. Firstly, thecorresponding error equations of the two kinds of quaternion EOPs models are derived. Thentwo kinds of quaternion bundle block adjustment models based on multi sensors are designedand established. At the end, the features of every adjustment model and the requirement ofGCPs are tested with actual high resolution satellite images. This method will provide a newtechnique for photogrammetry processing of the high resolution satellite images.