Abstract Details
153 / 2023-08-29 10:56:04
Retrieving high-precision three-dimensional deformation with three InSAR LOS measurements using L^2-norm minimization: application to the Wuhai coalmine area
InSAR, 3-D deformation, error allocation, L^2-norm minimization, coalmine
Abstract Pending
In the past two decades, interferometric synthetic aperture radar (InSAR) and multi-temporal InSAR (MT-InSAR) have been successful monitoring techniques for geological disasters such as landslides, earthquakes, volcanic eruptions, mining subsidence, etc. InSAR provides several advantages over GPS measurements, including high spatial resolution, large-scale continuous coverage, low cost, and accurate monitoring within a few millimeters. Nonetheless, since InSAR only gives one-dimensional deformation in the line of sight (LOS) direction, LOS deformation cannot quantify the scope of disasters in most practical applications. More than three components along the LOS direction, which should not be in the same plane, such as at different heading and incidence angles, can be used to construct an equation system to solve for three unknowns of vertical, east, and north deformation easily. However, deformation accuracy in the north direction is significantly poorer than in the vertical and east directions. The primary reason is that the analytic formula of the 3-D deformation solution has ill-conditioned problems with a significant error allocation coefficient from LOS to the north direction. Nevertheless, SAR focusing and InSAR workflow processing would inevitably bring errors from several sources, including decorrelation noise, atmospheric propagation delay, phase-unwrapping errors, DEM inaccuracy, etc., and these errors will lead to severe contamination for the north deformation.
Combining LOS and azimuth measurements, which can be estimated by InSAR, offset tracking, multiple-aperture interferometry (MAI), or burst overlap interferometry (BOI) techniques, could retrieve the 3-D deformation based on acquisitions from the ascending and descending tracks. Moreover, high-precision azimuth deformation estimated by MAI or BOI is highly dependent on imaging modes, particularly for the overlapping area covered by forward and backward imaging, which necessitates high doppler bandwidth and pairwise SAR coherence for the study area. However, most InSAR satellites cannot currently meet this strict requirement. With the launch of upcoming InSAR satellite missions following more frequent acquisitions, such as NISAR, Sentinel-1C, ALOS-4, and ROSE-L, researchers will be able to resolve small-level time-varying deformation in the same study area with more than two InSAR tracks. In addition, a single InSAR satellite, such as Sentinel-1A&B, might alter its orbit to get three images of the area of interest (Ascending and descending tracks plus the other track). This is crucial for rapidly acquiring the 3-D surface deformation of disaster areas such as earthquakes.
The current method could obtain vertical and east deformations with high accuracy. In contrast, the north deformations are less accurate and more difficult to solve due to the significant error allocation coefficient and ill-conditioned problems. Tikhonov regularization referred as L2
-norm minimization is often used to solve ill-conditioned problems. This paper proposes an efficient method for estimating 3-D deformation using only three InSAR LOS measurements and L2-norm minimization (referred to as TLOS-InSAR), especially for improving the precision in the north direction, without considering the imaging modes such as Stripmap, Spotlight map, ScanSAR, or TOPS, and applying complex signal processing methods.
We demonstrate the precision of our method by comparing our solution to GNSS in time series using Sentinal-1A three-track InSAR data covering the Wuhai (China) coalmine area in 2020-2021. Compared with Case 1 (resolving 3-D deformation considering the north one), our proposed method improves the accuracy of north deformation by 94%. The deformation precision in the vertical and east directions is nearly identical to Case 2 (resolving 2-D deformation neglecting the north one). The RMSEs in the 3-D direction are 6 mm in the vertical direction, 6 mm in the east direction, and 12 mm in the north direction. Specifically, the precision of north deformation has increased by 94% compared to the previous method.
Combining LOS and azimuth measurements, which can be estimated by InSAR, offset tracking, multiple-aperture interferometry (MAI), or burst overlap interferometry (BOI) techniques, could retrieve the 3-D deformation based on acquisitions from the ascending and descending tracks. Moreover, high-precision azimuth deformation estimated by MAI or BOI is highly dependent on imaging modes, particularly for the overlapping area covered by forward and backward imaging, which necessitates high doppler bandwidth and pairwise SAR coherence for the study area. However, most InSAR satellites cannot currently meet this strict requirement. With the launch of upcoming InSAR satellite missions following more frequent acquisitions, such as NISAR, Sentinel-1C, ALOS-4, and ROSE-L, researchers will be able to resolve small-level time-varying deformation in the same study area with more than two InSAR tracks. In addition, a single InSAR satellite, such as Sentinel-1A&B, might alter its orbit to get three images of the area of interest (Ascending and descending tracks plus the other track). This is crucial for rapidly acquiring the 3-D surface deformation of disaster areas such as earthquakes.
The current method could obtain vertical and east deformations with high accuracy. In contrast, the north deformations are less accurate and more difficult to solve due to the significant error allocation coefficient and ill-conditioned problems. Tikhonov regularization referred as L2
-norm minimization is often used to solve ill-conditioned problems. This paper proposes an efficient method for estimating 3-D deformation using only three InSAR LOS measurements and L2-norm minimization (referred to as TLOS-InSAR), especially for improving the precision in the north direction, without considering the imaging modes such as Stripmap, Spotlight map, ScanSAR, or TOPS, and applying complex signal processing methods. We demonstrate the precision of our method by comparing our solution to GNSS in time series using Sentinal-1A three-track InSAR data covering the Wuhai (China) coalmine area in 2020-2021. Compared with Case 1 (resolving 3-D deformation considering the north one), our proposed method improves the accuracy of north deformation by 94%. The deformation precision in the vertical and east directions is nearly identical to Case 2 (resolving 2-D deformation neglecting the north one). The RMSEs in the 3-D direction are 6 mm in the vertical direction, 6 mm in the east direction, and 12 mm in the north direction. Specifically, the precision of north deformation has increased by 94% compared to the previous method.
Important Date
-
Conference Date
Oct 26
2023
toOct 29
2023
-
Oct 15 2023
Abstract Submission Deadline
-
Oct 15 2023
Draft paper submission deadline
-
Nov 13 2023
Registration deadline
Sponsored By
International Society for Mine Surveying
China Coal Society
China Surveying and Mapping Society
China Coal Society
China Surveying and Mapping Society
Organized By
中国矿业大学
中国煤炭科工集团有限公司
中国煤炭科工集团有限公司
Contact Information
