Imaging Models And Geo-positioning Technology For High Resolution Satellite Imagery

As spacial information processing technology of high resolution remote sensing images is one of the important research areas of numerical photogrammetry, geographic information system and other related disciplines, this paper systemically discusses the rigorous sensor model, non-regorous sensor model, rational function model and model-based geo-position technology. Transformation between the rigorous sensor model and rational function model is also investigated. The work accomplished in this paper is as follows:1. The arithmetic of space resection to calculate the orientation parameters, methods to overcome the relativity between the parameters and space intersection of satellite images are introduced. Experiments including reading and parsing the QuickBird image support files to compute the rigorous sensor model, geo-positioning, and analysis the positioning accuracy and the regularity of error distribution, are conducted. The positioning error is about 20m, which is consistant with the declared accuracy of 23m.2. The applicability of several kinds of non-rigorous imaging model for QuickBird imagery is investigated, and time-variant parallel perspective model is proposed. Through comparison of the accuracy in image space and object space, time-variant parallel perspective model is considered as the best model to describe the imaging geometry of QuickBird imagery, and the geo-positioning accuracy can reach 0.6m in planimetry and 0.8m in height.3. The rational function (RF) model and its downward model are introduced. Arithmetics of parameter calculation and stereo-position are derived. Test of parameter calculation is conducted using supplied rigorous model and known control points. The results show that the precision of terrain-dependent scheme is low while the terrain-independent scheme has a high one. By using the rational polynomial coefficients supplied by the QuickBird imagery, stereo-position experiment is conducted and several types of compensating models are proposed based on error distribution analysis. The results show that, the positioning accuracy can be enhanced to 1m by using only 2 well distributed control points, and it can reach 0.6m to 0.8m in planimetry when 5 to 6 well distributed control points are available by using the affine model. If the control points are adequately enough (n≥6 and well distributed), we can get a higher accuracy of 0.6m in planimetry and 0.8m in height by using the 2-order polynomial model in image space.4. As rigorous sensor model and rational function model each has advantage and disadvantage, the conversion of the two models is investigated. The results indicate that rigorous sensor model can be recovered from rational function model to a certian extent. Using the virtual control points generated by the rigorous sensor model, RF coefficients can be computed. The derived rational function model has the approximate accuracy with the rigorous sensor model, with about±lmm error in stereo positioning. Since the 10 parameters x₀ and f are correlated for linear array images, x₀ should be fixed to recover the rigorous sensor model. If x₀ is set to the true value, the rigorous sensor model can be retrieved with high accuracy of±0.001mm error in internal orientation parameters,±5cm error in exposure center coordinates and±0.002" error in rotation angles. The specification of x₀ only causes a small rotation between the real camera coordinate system and the specified camera coordinate system, the exposure center coordinates and the stereo positioning accuracy are theoretically not affected. By using the recovered rigorous sensor model, stereo positioning accuracy decline is within 1mm.In the finality, the work which has been done is summarized, and the problems requiring further research are discussed.This paper is supported by National Natural Science Foundation of China (Grant No. 40771174), National High Technology Research and Development Program of China (Grant No. 2007AA12Z178) and SHUGUANG Project (Grant No. 07SG24).