Stereo vs. tri-stereo: Preliminary results on Anthropogenic-induced landform change detection using satellite derived remote sensing data

Abstract

The Trans-Sulawesi railway construction runs for 144 km from South to North of the South Sulawesi's east coast. The railway line mostly was built on an embankment over agricultural flat, while on the hill, the railway line built in an incised slope. These actions essentially alter the surrounding landform. The changing landform could have its effect on how the surface runoff flow during the rainy season. In this study, we explore the potential of satellite-derived digital elevation models (DEMs) on detecting the changed landform over the Takkalasi Watershed. We employed the SPOT-7's tri-stereo acquisition mode, acquired on 17th April 2020. In these preliminary results, we compare the DEM results from each pair combination as well as the combination of all three pairs. The use of different tie-points (TP) and ground control points (GCP) were also investigated. Fast Fourier Phase Transformation algorithm were employed for the TP extraction, while the semi-global matching algorithm were employed for the epipolar DEM extractions. The produced DEMs then validated using the reference data taken from dGPS, UAV DEM, as well as the ICESat-2 ATL08 data. Our initial results shows that the best DEM generated is produced by the forward and backward stereo pair using only nine (9) GCP with the average vertical Root Mean Square Error (RMSE) of 4.4 meter. At 3 m ground sampling distance, the incised hill is easier to detect on SPOT-7 DEM compared to the embankment.