SPOT-5/HRS STEREO IMAGES ORIENTATION AND AUTOMATED DSM GENERATION

摘要

HRS (High Resolution Sensor), carried on SPOT-5, is the first high-resolution sensor on the SPOT constellation that enables the acquisition of stereo images in pushbroom mode from two different directions along the trajectory. The Institute of Geodesy and Photogrammetry (IGP) participated as Co-Investor in the ISPRS-CNES initiative for the investigation on DEM generation from SPOT-5/HRS stereoscenes. This paper describes the work carried out at IGP on a stereopair acquired on 1st October 2002 over parts of Bavaria and Austria. For orienting HRS imagery, two alternative approaches have been used: a rigorous sensor model and a rational function model. Both algorithms have been implemented at our Institute. The rigorous sensor model is based on the classical collinearity equations, which are extended by the sensor external orientation modeling with 2nd order piecewise polynomials depending on time and by selfcalibration parameters. Using well distributed Ground Control Points (GCPs), the unknown internal and external parameters are estimated with a least squares solution. The alternative method, independent from the camera model, does not describe the physical imaging process, but uses rational polynomials to relate image and ground coordinates to each other. This algorithm consists of two steps: 1) calculation of Rational Polynomial Coefficients (RPC) for each image with a least-squares using the geometric information contained in the metadata files; 2) block adjustment with the computed RPC model using GCPs. Both orientation methods gave RMS errors in Check Points (CPs) in the range of 1 pixel in all coordinate directions. Using the triangulated orientation elements, the DSM was extracted with algorithms and software packages for CCD linear sensors developed at IGP. After the creation of image pyramids, the matches of three kinds of features (feature points, grid points and edges) on the original images are found progressively in the next levels starting from the low-density features on the images with the lowest resolution. A triangular irregular network based DSM is constructed from the matched points on each level of the pyramid and used in turn in the subsequent pyramid level for the approximations and adaptive computation of the matching parameters. A modified Multi Photo Geometrically Constrained Matching algorithm is employed in order to achieve sub-pixel accuracy for all the matched features. The extracted DSM was compared to the reference DSMs obtained from laser data and map contours at different spacings (5X5 m, 25X25 m and 50X50 m) provided by DLR in Oberpfaffenhofen using both terrain height and orthogonal distances. The results show RMS values between one and two pixels on the average and a systematic error mainly due to the presence of trees. After a manual removal of the main areas covered by trees in the reference DSMs sites those errors have been removed. The final results show a mean error in the range of 1-5 meters.