Interferometric Synthetic Aperture Radar (InSAR) is an established technique to extract 3D information of surfaces by using SAR images. InSAR techniques provide the capability to detect surface deformation and to generate Digital Elevation Models (DEM). However, while the methodology is well-established, it remains challenging to produce convincing InSAR DEM over vegetated areas due to temporal decorrelation. In order to address the issue, we aim at improving the quality of InSAR DEM by selecting appropriate datasets in the pre-processing stage, and by systematically assessing parameter variations during processing, especially for vegetated areas. Using C-band Sentinel-1A and Sentinel-1B satellite images, the test has been conducted over the Taichung area of Taiwan, where images cover both urban and vegetated areas. Winter and summer datasets are applied to observe seasonal effects. Besides, a multi-look processing procedure is carried out to inspect the influence of this process. Our experimental results demonstrate that winter datasets outperform summer datasets, achieving up to 13 meters of Root Mean Square Error (RMSE) in urban areas. It is also noted that multi-look processing procedure has different effects in our results; it improves the result for the summer pair but reduces the accuracy for the winter pair. In addition, one common property for both seasonal effects and multi-look processing effects is that the processing has more evident impact over higher elevated vegetated terrain than plane urban areas. To conclude, selecting datasets of appropriate season, and deciding if applying a multi-look processing procedure based on the property of datasets could improve the reliability of InSAR DEMS.
|已出版 - 2020
|40th Asian Conference on Remote Sensing: Progress of Remote Sensing Technology for Smart Future, ACRS 2019 - Daejeon, Korea, Republic of
持續時間: 14 10月 2019 → 18 10月 2019
|40th Asian Conference on Remote Sensing: Progress of Remote Sensing Technology for Smart Future, ACRS 2019
|Korea, Republic of
|14/10/19 → 18/10/19