Urban Atlas – DLR Processing Chain for Orthorectification of Prism and AVNIR-2 Images and TerraSAR-X as possible GCP Source

Kurzfassung

In the first part of this paper, the semi-automatic processing chain established at the DLR for the processing of PRISM and AVNIR-2 data in the framework of Urban Atlas is described. Although the processing chain was designed to be fully automatic, the lack of a sufficient accurate reference image necessitates the manual measurement of ground control points (GCP) from digital maps and aerial imagery to fulfill the accuracy requirements of 5 m RMSE in case of PRISM images. In case of AVNIR-2 images, the internal reference database available at DLR was sufficient accurate to allow for an automatic GCP generation. Since no ground truth was available, the included quality check uses different criteria to allow for a quality measure for each image. In the second part of the paper, the accuracy of the orthorectified scenes is analyzed. Up to now, more than 200 PRISM and 100 AVNIR-2 scenes for 43 European cities are processed. The overall accuracy statistics are presented and additionally some single scenes are evaluated in areas where additional ground truth is available. In further investigations, TerraSAR-X data was evaluated regarding its potential as source for GCPs. The very high geometric accuracy of geocoded data of the TerraSAR-X satellite has been shown in several investigations and thus qualifies the data as GCP source. Different methods of retrieving GCPs from TerraSAR-X data have been evaluated in the third part of the paper: manual measurements or local image matching using mutual information. By adjustment calculations, falsely matched points can be eliminated and an optimal improvement can be found. After the orthorectification of the PRISM data using these improvements, the results are compared to PRISM data that were orthorectified using conventional ground control information from GPS measurements.