PSI - Issue 64

H.F. Stewart et al. / Procedia Structural Integrity 64 (2024) 573–579 Stewart, Cusson, Greene Gondi & Oliver / Structural Integrity Procedia 00 (2024) 000–000

577

5

2.3. Satellite radar image acquisition and processing PS-InSAR analysis was conducted on high-resolution Spotlight imagery from the C-Band RADARSAT-2 satellite, which operates in a sun-synchronous orbit at an altitude of 798 km. A total of 26 scenes from the SLA12-D descending stack were selected for the PSI analysis (MDA 2021). These images were acquired from 21 June 2019 to 4 July 2021 and had an incidence angle of 39°, ground range resolution of 2.5 m, ground azimuth resolution of 0.8 m, and a repeat pass cycle of 24 days. This viewing geometry is preferred due to its high sensitivity to deformations in the longitudinal direction of the bridge – the direction in which large thermal deformations can be observed. Radar scenes were corrected for ambient air temperature effects using measurements from an in-situ weather station at the Montreal–St. Hubert Airport. More details can be found elsewhere (Cusson and Stewart, 2024). SAR images have a side-looking geometry, which may cause viewing distortions such as foreshortening and layover that displace features from their true positions by an amount proportional to their height. To account for this, a Digital Elevation Model (DEM) is used during the georeferencing process. The PS-InSAR analysis and georeferencing used during the validation trial of the RS-SHM method was performed using a single tile of the Shuttle Radar Topography Mission (SRTM) DEM, which is a 1 arc-second SRTM DEM (30 m resolution) acquired in 2000. Horizontal positions of the SRTM DEM are given in WGS84 geographical coordinates, and heights are given relative to the EGM96 geoid model. The SRTM DEM represents only the bare-earth terrain and does not contain any data from the bridge or nearby buildings, so as part of the InSAR analysis, a height error analysis was performed. The height difference between each PS target and the SRTM heights was estimated using a least squares linear phase model fit to the network of unwrapped phases. This was then added to the SRTM elevations to give the estimated elevation of each PS target. Non-constant offsets were observed between the positions of visually distinctive PS InSAR targets along the bridge and their positions observed in orthorectified optical imagery. These offsets varied along the horizontal longitudinal direction of the bridge as the height and orientation of the bridge changed, and target heights above the bridge deck were also observed to vary. An Airbus Pleiades 1B orthorectified image of the bridge acquired on 22 June 2020 was used to manually correct these offsets by correlating the positions of PS targets to the positions of specific bridge elements visible in the image, and the offsets between them were then used to generate warping polynomials to adjust the positions of the remaining PS targets. To assess whether the Pleiades DSM offers any improvements to the geolocation accuracy of PS targets, the georeferencing procedure was repeated using the Pleiades DSM. The clipped Pleiades DSM provided for PS-InSAR analysis only contained the bridge structure, but georeferencing requires elevation information for all PS targets. Therefore, the Pleiades DSM was merged with the SRTM DEM. To do this, the elevation models were resampled to a common pixel spacing of 1 m and transformed from WGS84 ellipsoid elevations to EGM96 geoid heights, NoData values were assigned, and a minor boxcar filter was applied to smooth the edges between models. The refined height corrections generated using the SRTM DEM during the PS-InSAR analysis were applied to the merged DSM in areas in which the clipped Pleiades DSM had no coverage to the north and south of the bridge. 3. Results and analysis Vertical error values on the test bridges were all found to be within the expected error tolerances, and no bridge had a vertical RMS error exceeding 1 m. The results are shown in Table 1.

Table 1. DSM error comparison results over selected bridge structures. Structure RMSE Z (m)

Accuracy LE 90 (m) Average height (m)

No. of samples

Champlain Bridge (decommissioned)

0.49 0.66 0.39 0.42 0.78 0.64

0.81 1.09 0.64 0.69 1.29 1.06

28.97 20.02 21.20 22.56 43.04 20.63

69 30 23 54 76 23

Clément Bridge

Estacade ice control structure

Concorde Bridge (main span between islands) Jacques Cartier Bridge (east of main span) Victoria Bridge (east of Saint-Lambert Locks)