21st Congress of the International Society for Photogrammetry and Remote Sensing, ISPRS 2008, Beijing, Çin, 3 - 11 Temmuz 2008, cilt.37, ss.865-871
© 2008 International Society for Photogrammetry and Remote Sensing. All rights reserved.For an extension of the water catchments areas for the Greater Istanbul Municipality investigations in a mountainous forest area in Turkey, at the Bulgarian border, named Istiranca, have been made. In the project area a height model from SPOT-5 HRS, the SRTM C-band height model and a digital elevation model (DEM) for the topographic map 1:25 000 are available like also a Cartosat-1 stereo pair. For comparison a SPOT-5 HRS height model has been checked in Istanbul vicinity against a reference DEM from more accurate maps 1:5000 and also the SRTM C-band height model.Based on the Cartosat-1 stereo pair a digital surface model (DSM) has been generated by automatic image matching. For the orientation of the Cartosat-1 images control points from the topographic map 1:25 000 have been used, leading in the average to RMSX=6.78m, RMSY=7.11m and RMSZ=6.08m. This accuracy is dominated by the limited control point quality, but with the 14 well distributed control points a satisfying orientation can be guaranteed. The automatic matching by least squares resulted in an astonishing complete coverage of the area by matched points. The matching failed only in small sub-areas covered by clouds. In the forest, not influenced by clouds, more than 98% of the possible object points have been matched successfully with correlation coefficients above 0.5 and with a maximum of the correlation coefficients in the range of 0.85. For mountainous forest areas this is an unusual good result. The matching was made for every third pixel, leading to a point spacing of approximately 7.5m. SRTM C-band and the results from image matching with optical images lead to DSMs, including the height of the visible surface, while the reference height model from the topographic map 1:25 000 refers to the bare ground. The SRTM height model is available with a spacing of 3 arcsec, including only limited morphologic details. All the height models have been compared to each other. A strange correspondence of the SPOT-5 HRS height model to the SRTM Cband height model can be seen, which only can be explained by filling the gaps in the SPOT-5 HRS height model in the forest area by SRTM C-band data.