Analysis of Cartosat-1 images regarding image quality, 3D point measurement and DSM generation

The Institute of Geodesy and Photogrammetry (IGP) at ETH Zurich is participating in the Cartosat-1 evaluation programme, a common initiative of ISRO (India) and ISPRS. Within this programme various test sites with reference data have been established and Cartosat-1 images have been acquired over these sites. Here, investigations at the Rome (Italy) and Maussane-les-Alpilles (France) test sites are reported. First, radiometric problems encountered with the images and pre-processing for their improvement are reported. Examples are shown, together with pre-processing methods that can be employed in order to improve image quality, aimed especially at automatic generation of a digital surface model (DSM) with fewer blunders and more matched points. Then, georeferencing is discussed and the 3D point measurement accuracy that can be achieved is introduced, as well as problems encountered with the rational polynomial coefficients (RPCs). The georeferencing results were produced using various options regarding image pre-processing, the mathematical model used for georeferencing, the number and distribution of ground control points (GCPs) and the GCP image mensuration methods. The best results led to a planimetric and height accuracy (RMSE) of about 1.3 m. Use of an affine transformation after the RPCs, with about six well-distributed GCPs transferred with matching to the second image, is suggested. Finally, the results of automatic DSM generation using the SAT-PP program package developed at the IGP are presented. Various DSMs were generated with 10 m grid spacing. The results were checked visually and were also compared to the reference data provided. In the best case, the accuracy achieved is about 2.7 m without any manual editing, in spite of a 3-year difference between the matching and reference DSMs. Although some aspects regarding image quality and RPC generation could be improved, Cartosat-1 is a useful sensor for mapping and especially for the generation of DSMs. However, owing to its poor absolute geolocation accuracy, Cartosat-1 cannot be used for the generation of a global DSM without GCPs.