Methodology and Techniques - PSI

Cultural heritage is a key element of history, because the ancient monuments and archaeological sites enrich today’s societies and help connect us to our cultural origins. It is also an important resource for economic growth, employment and social cohesion, offering the potential to revitalize urban and rural areas and promote sustainable tourism. The diverse historical sites located in urban or rural areas are susceptible to displacement of Earth’s surface as well as other motion anomalies, which could lead to dislocation of structures. Protection and conservation of our cultural assets for future generations in the face of various natural or anthropogenic hazards is a major concern nowadays. Understanding the deformation process is essential for interpreting the archaeological finds and the site formation and of course to protect them from the related hazards.

During the last years, spaceborne Synthetic Aperture Radar (SAR) interferometry, particularly the innovative Multi-Temporal SAR interferometry (MT-InSAR), has proven to be a powerful remote sensing, non-invasive, tool for detecting and measuring ground deformation and studying the deformation’s impact on man-made structures.

Under the framework of Sino-Greece bilateral cooperation, this study aims to demonstrate first results in cultural heritage conservation contributed by the collaborative investigations: that is, Persistent Scatterers SAR Interferometry (PSInSAR) to monitor potential ground deformation affecting the archaeological remains in the great importance historical center of Athens (Greece) and Tomography-based PSInSAR (Tomo-PSInSAR) to monitor regional-to-monument scale displacements to figure out the impacts of urbanization, material thermodynamics and local geology for the sustainable development of ancient city walls  (dated back to 600 B.P.) in Nanjing (China). For this purpose, two datasets including 52 scenes of Sentinel-1 IW images (ascending with VV polarization) and 20 scenes of TerraSAR-X stripmap images (descending with HH polarization), spanning the periods 12/10/2014 - 26/08/2016 and 5/3/2012 - 30/08/2016 respectively, were applied for the Athens exploitation; and 26 scenes of TerraSAR/TanDEM-X strimap images acquired in the span from May 2013 to February 2015 were applied for the ancient city wall (Nanjing) monitoring.

Major objective for this research, aside from broadening our fundamental understanding of the combined impacts of Earth’s potential deformation, material thermodynamics and anthropogenic activities on archaeological sites, assessment the capability of Sentinel-1 and TearraSAR/TanDEM-X with varying scale, wavelength and polarization to monitor vulnerability over monuments is also significant for the promotion of MT-InSAR tools in the safeguarding and conservation of cultural heritage sites. Moreover, the proposed Tomo-PSInSAR facilitates structural instability monitoring of monuments, particularly for those with multi-terraces in vertical, owing to its capability for extraction overlaid PSs. 

The primary bottlenecks that hinder the widespread use of Differential Synthetic Aperture Radar Interferometry (DInSAR) technique for bridge Structural Health Monitoring (SHM) are the difficulties in selecting dense Point Targets (PTs), modeling thermal dilations, investigating irreversible abnormal behaviors in full scale, and rendering results in 3D view. To address these challenges, we developed a meticulous and reliable multi-temporal DInSAR approach that can retrieve and analysis the regular pattern of thermal dilations and subsidence with an enhanced density and accuracy of PTs by considering the bridge structural characteristics.

In terms of complex bridge structures, the inherently oblique scene illumination may cause ghost signals in SAR images. Besides, they are vulnerable to suffer from decorrelation problem on account of the vibration under environment effects. Therefore, the identification of PTs on the full-scale of bridges remains a significant challenge compared to other manmade objects. Here, the density and precision of measurements are improved by considering the back-scattering characteristics and vibration of bridges. Theoretically, double-bounce and triple-bounce stripes associated with bridges are both virtual objects, while only single-bounce points are useful for deformation detection. A multi-temporal strategy effectively combines improved Persistent Scatterer Interferometry (PSI) and Small Baseline (SBAS) method is implemented to maximum the spatial density of single-bounce PTs detected. Firstly, the amplitude dispersion, coherence and priori information of structures are combined to select the dominant scatterers on the outline of the bridge. Then, the spectrally-filtered interferograms with small baselines are used to minimize the decorrelation so that the semi-stable scatterers on the bridge deck can also be detected. Regarding the vibration of bridges, a higher maximum topography error tolerance over water surface is set for PTs identification.

Regarding thermal dilations, a quantitative exploration of seasonal temperature variation effects is performed for detecting the nonlinear deformation. This temperature-dependent components separation is considered in post-processing and interpretation so that it will not affect the previous estimation. A regression-based thermal response model is established by linear fitting between the residual interferometric phase signal obtained by iterative spatial-temporal filtering and temperature differences. With the assumption that temperature is homogeneous along the structure, seasonal thermal dilation effects and thermal coefficient parameter of the structure can be accurately estimated. As bridges are hyper-static structure, the presence of piers may hinder the accumulation of thermal dilation along the structures, which resulting the maximum accumulation thermal dilations occurs in the central part. The actual physical property of material can be used to verify the estimated thermal coefficient parameter.

According to the distribution and characteristics of coherent targets, bridges can be classified into two major types: coherent and partially coherent. For coherent bridges, sufficient PTs can be identified for a full-scale deformation investigation. For partially coherent bridges, we developed a Deformation-Sensitive Feature Points (DSFPs) modeling and analytical method to enable record and interpretation the global structure behaviors. The response and mechanical characteristics of structure, as well as the distribution and deformation of PTs by InSAR are considered as selection criteria of DSFPs. Then, the locations, types and stabilities of DSFPs are analyzed for damage assessment. The results are verified by in-situ measurements or by cross-validation. In addition, a novel authentication method based on structural reliability theory was proposed by comparing DInSAR results with the structural mechanics principle.

3D visualization can benefit for understanding of InSAR measurements in practical applications. In this step, the key idea is to establish the optimal relationship between SAR pixel coordinates and the corresponding geographic coordinates. Here we employed the Rational Polynomial Coefficients (RPC) Model to implement the transformation between 3D geographic coordinates and 2D images coordinates of PTs identified on the bridges. After the coordinate ortho-rectification, the geographically-correct 3D bridges are reconstructed.

To demonstrate the effectiveness of our approach for bridge-specific SHM, case studies on different types of bridges in Shanghai, Tianjin and Wuhan based on high-resolution TerraSAR-X, Cosmo-SkyMed and medium-resolution Sentinel are carried out. Specially, comparison between results from X-band and C-band SAR images is conducted to show the weakness and strength of each datasets. Our results suggest that a SHM system based on multi-temporal DInSAR measurements and improved analytical model, if well designed and implemented, can not only detect the meticulous and reliable bridge deformation patterns and magnitudes, but also improve future designs and facilitate diagnosis pre- and post-hazard conditions. 

At Fringe 2009 Fabrizio Novali introduced SqueeSAR, a refined approach to deformation analysis developed by Ferretti, Fumagalli, Novali himself, Prati, Rocca and Rucci. The presented comparisons of results generated with SqueeSAR to such generated with PSInSAR exhibited an impressive increase of extracted information. This success is based on a unified spatio-temporal statistical analysis which can be seen to turn DS (distributed scatterers) into PS. Means of this conversion is a maximum likelihood estimator, derived and theoretically investigated by Monti Guarnieri and Tebaldini (2008), that provides the complex signal common to a group of DS pixels obeying a joint circular complex normal distribution. In the framework of SqueeSAR, published 2011 in TGRS, this estimator is made applicable by providing ideas for grouping DS candidate pixels in a neighborhood of an investigated pixel, which is supposed to be statistically homogeneous and to allow for the estimation of the covariance matrix.

Despite the demonstrated capability of this approach to enhance coverage with high quality information considerably, questions which arise naturally when one tries to put it to work were not answered satisfyingly. The key point is that real data rarely fulfill the assumption of joint circular complex normal distribution in an ideal way. At Fringe 2015, we therefore introduced an augmented statistical model for DS, that can be used for the development of methods for estimating the DS phase history when the scale of backscattering is varying, residual fringes are present and a limited number of outliers contaminate otherwise Gaussian distributed data. Furthermore, we showed that this approach extends to complex elliptically symmetric distributed random vectors in case the density generator is strictly monotonically decreasing. Since then, we investigated many possible choices of sub algorithms, estimators and parameter settings with respect to their performance (intermediate results were presented at EUSAR 2016). In this work we continue these investigations with emphasis on ways of considering residual fringes for the estimation.

For the purpose of exploring these questions we implemented a testing environment. DS candidates are processed in a detail of the investigated scene and then connected to a reference result of a PS analysis. Pixels are assessed according to the temporal coherence of their phase differences to nearby PS and versus ground truth. Test data are from a scene in Bavaria and a desert scene, where LIDAR DEMs are available and from the town of Lüneburg, Germany, where we use deformation series from leveling as ground truth. All stacks are TerraSAR-X high resolution spotlight-mode (300MHz).

Ferretti, A., Fumagali, A., Novali, F., Prati, C., Rocca, F., Rucci, A. (2009): “The Second Generation PSInSAR Approach: SqueeSAR,” Fringe 2009, ESA-ESRIN, Frascati.

Ferretti, A., Fumagali, A., Novali, F., Prati, C., Rocca, F., Rucci, A. (2011): “A New Algorithm for Processing Interferometric Data Stacks: SqueeSAR,” IEEE Transactions on Geoscience and Remote Sensing, Vol. 49, Issue 9.

Monti Guarnieri, A., and Tebaldini, S. (2008), “On the exploitation of targets statistics for SAR interferometry applications,” IEEE Trans. Geosci. Remote Sens., Vol. 46, No. 11, pp. 3436–3443, Nov. 2008.

Even, M. (2015), “Advanced InSAR processing in the footsteps of SqueeSAR,” Fringe 2015, ESA-ESRIN, Frascati.

Even, M. (2016), “A study on spatio-temporal filtering in the spirit of SqueeSAR,” EUSAR 2016, Hamburg.

The potential of SAR interferometry technology to detect and monitor ground and structure deformations has been fully demonstrated and now this technology has become an operative instrument for mapping territory in order to prevent geo-hazards phenomena. Nowadays there are several sensors with complementarity characteristics, in terms of spatial resolution and tasking flexibility, which allow interferometric analyses at various levels, thus optimizing the use of resources and increasing the efficiency of the overall analysis process.

The ESA Sentinel-1 mission, specifically designed to collect interferometric SAR data, is opening new challenges and opportunities, making possible to monitor surface deformations worldwide and routinely, thanks to the capability of performing frequent acquisitions of very large areas in the interferometric wide-swath TOPS mode, even though with a medium-low spatial resolution.

In this context, high resolution X-band SAR systems like COSMO-SkyMed or TerraSAR-X can be complementarily exploited for detailed analysis of restricted areas, chosen based on a particular interest, or also based on the lower resolution analysis performed with Sentinel-1 data. Moreover, X-band systems in principle guarantee a higher sensitivity to displacements than C-band sensor like Sentinel-1.

In the work, we will quantitatively discuss the complementarity of Sentinel-1 and COSMO-SkyMed constellations, both with theoretical analyses and experiments based on interferometric deformation measurements obtained by persistent scatterer PS SAR interferometry processing of data from the two systems. In particular, in addition to the known characteristics like swath extension, spatial resolution, revisit time, we will compare:

- Phase noise levels, and the resulting PS deformation measurement accuracy

- PS densities both on urban and natural terrains

- PS 3D positioning accuracy, depending not only on the SAR spatial resolution, but also on the phase noise and the baseline values.

We will then report different examples showing the complementary characteristics in practical use cases for different applications. As a conclusion, we will also envisage different possible operational scenarios to take the best from the medium-low and high resolution systems, based on the complementary characteristics.

3.02.b RT: Methodology and Techniques - PSI
Location: Nokia-Sali