Comparison of optic and radar data for terrain feature extraction

The availability of suitable maps of terrain features is very important because these complications and the process governing them are the sources of many environmental hazards and resources. The purpose of this research, compare the optic and radar data for terrain feature extraction. A panchromatic band of Landsat 8 (OLI) as an optical image and two C-band Sentinel-1 satellite radar images with a resolution of 22×20 m per pixel were used. Two radar images (radar-1 and 2) in two different acquisition directions with different look angels were visually interpreted. In order to reduce the geometric and radiometric effects of the topography, the radar images were corrected using 3-second SRTM data (modified radar). In this study, four feature types were extracted by visual interpretation, which is: valleys, blades, alluvial fans, and debris fans and the results were compared with the World Imagery layer. The accuracy, completeness, quality, kappa, and z-test were calculated for every interpretation. The results showed that the highest accuracies in recognition of valleys and blades using Landsat image were 83.90% and 87.88%, respectively, and the highest accuracies of alluvial fan and the debris fan of the modified radar image were 82.76% and 83.72% respectively. The highest kappa coefficient related to the modified radar data was calculated at 54.72% (Landsat 49.74%) and the highest z-text related to Landsat-radar 1 was calculated 0.9871 (Modified Radar-Landsat 0.6443%).