Three dimensional data sets were used to produce and evaluate Digital Elevation Models (DEMs) from High Resolution Satellite Images (HRSI). The main objective of this research is to test and assess the use of HRS SPOT5 in DEM production. Different DEMs have been generated for testing the effect of different processing parameters on their accuracy. The radiometric variations of the images is reduced due to that the data of the used stereo-pair images were acquired on the same date. Georeferencing and stereo model set up of the images have been performed by five different methods. These methods are: DEM production by using metadata only, by using tie points generated automatically, by using tie points generated manually, by using tie points generated automatically in the triangulation process with 9 GCPs (XYZ coordinates) and by using tie points generated automatically and manually. Orthoimage based on a reference DEM is produced for extracting GCPs. The results of the accuracy statistic of the different generated DEMs indicate that the elevation accuracy was better from zero to 3 meters in flat and even in mountainous and steep areas. The results also indicate that the DEMs generated with auto and manual tie points and that with auto tie and ground control points are close from that generated with metadata only. So, in the case of metadata absence, auto or manual tie points with GCPs can be used for DEM production with accuracy close from that achieved with using metadata only. The Root Mean Square (RMS) of errors within 95 % confidence level was found nearly to 2.77 m which is equivalent to 1.108 pixels in the case of using auto tie points and GCPs . This is considered a high accuracy. |
