MUZUBI – Multi-Zone Phase Unwrapping using Advanced Split-Band Interferometry
———-
DURATION: 2016-2018 (2 years)
PARTICIPANTS:
- Centre Spatial de Liège (D. De Rauw, L. Libert)
- Royal Museum for Central Africa (F. Kervyn, B. Smets)
- European Center for Geodynamics and Seismology/National Museum for Natural History, Luxembourg (N. d’Oreye)
- Comahue National University, Argentina (Y. Rubio)
- Canada Centre for Mapping and Earth Observation, Natural Resources Canada (S. Samsonov)
SUPPORT:
– Belgian Science Policy (Belspo)
– DLR project NTI_INSA6714
DESCRIPTION:
CONTEXT
The current MUZUBI project is aiming at developing a novel methodology to improve the phase unwrapping in SAR interferometry (InSAR). The methodology is based on the results of a former project (Vi-X). The innovative method to be developed here will be tested on two case studies in Argentina and DRC. Results will be compared with the results from a state-of-the-art method (MSBAS) currently used in the frame of a running project (RESIST) in order to assess and quantify the benefit of the proposed methodology.
In SAR interferometry (InSAR), phase unwrapping remains the bottleneck to get continuous phase measurements among separated coherent areas, which is required to compute the topographic or the relative displacement component. In addition to giving relative measurements, classical interferometry often leads to unconnected patches, each unwrapped independently. For each patch, ground truth is required to allow connecting them together and get continuous measurement across the scene.
Unfortunately, in general, these ground truth data are unavailable. Most recent SAR sensors use wide band signals to achieve metric range resolution. One can also take advantage of wide band to split it into sub-bands and generate several lower-resolution images, centred on slightly different frequencies, from a single acquisition. This Split Band process, also named Multi Chromatic Analysis (MCA), corresponds to performing a spectral analysis of SAR images. Split-Band SAR interferometry (SBInSAR) is based on spectral analysis performed on each image of an InSAR pair, yielding a stack of sub-band interferograms. Scatterers keeping a spectrally coherent behaviour in each sub-band interferogram show a phase that varies linearly with the carrier frequency, the slope being proportional to the absolute optical path difference. This potentially solves the problems of phase unwrapping on a pixel-per-pixel basis. Of prime importance, it might allow connecting independent zones, estimating the absolute phase on most coherent pixels of each patch, provided that the quired accuracy could be achieved.
The MUZUBI project aims at filling that gap toward a fully developed Split Band-assisted phase unwrapping processor for SAR interferometry using Multi-Chromatic Analysis. A former project (Vi-X) showed the high potential of the technique to perform absolute height/displacements measurement, connecting independently unwrapped zones. However, the spectral diversity of the TanDEM-X (TDX) data appeared to be insufficient to achieve the required precision because of the bistatic focusing procedure that limits the bandwidth to 100MHz. It is thus proposed to use among others, more advanced sensed data that will be made available through a recently approved proposal submitted to the TanDEM-X Science Phase AO. These exceptional data will be acquired in both stripmap and spotlight pursuit mode, granting both wide bandwidth (spectral diversity) and high coherence.
It is thus proposed to adapt existing processor to specific pursuit and spotlight TerraSAR-X (TDX) acquisition schemes and combine SBInSAR processor with classical phase unwrapping procedure in order to get absolute phase measurement on all coherent zones.
The developed technique will be applied to the study and monitoring of two active volcanic zones: the Nyiragongo/Nyamulagira (RDC) and the Copahue (Argentina). In the first case, it should allow getting connected displacement measurements on separated areas around the highly vegetated volcanoes. In the second case, known to be more challenging in terms of topography, it should allow to resolve the required topographic component.
STATE OF THE ART
The proposed research aims at exploring and valorising this new degree of freedom offered by the wide range spectral band of high resolution SAR data to perform absolute ranging.
Azimuth band splitting is widely used since the very start of SAR imagery. It is used to generate several independent looks of a single acquisition to allow reducing speckle by stack averaging at the expense of a resolution loss [1]. It was also proposed by Arnaud in the frame of vessel detection to generate two looks between which coherence may be estimated; coherence being preserved on man-made structures while lost on sea clutter [2]. Azimuth band splitting is also proposed in the frame of Multi Aperture Interferometry, generating a forward and a backward interferogram from a given interferometric pair. In this way, displacements can be measured along different squints allowing to estimate the azimuthal component of displacements [3].
Range band splitting in two bands was proposed by Cumming & al. in the Multi-Look Cross Correlation (MLCC) technique to resolve Doppler ambiguity in the frame of Doppler centroid estimation [4]. Based on an original idea proposed by Madsen and Zebker [5], a novel SAR processing approach introduced in [6,7] was developed into the so-called multi-chromatic InSAR processing, involving the computation of a stack of equal-resolution interferograms from a single interferometric pair, each one being centred at a different carrier frequencies. The complex signal of each image pixel can then be studied as a function of the carrier frequency across the stack. The phase of suitable “frequency-coherent” scatterers evolves linearly with the sub-band central wavelength, the slope being proportional to the absolute optical path difference. This potentially solves the problems of phase unwrapping.
Recently, Split Band InSAR processing was also proposed to retrieve the ionospheric phase component in ALOS2 L-band interferograms [8]. In this case, split band interferograms are unwrapped classically to allow retrieving the ionospheric component through interferograms differentiation. In the frame of the WiMCA ESA project (contract N° 21319/07/NL/HE), CSL demonstrated feasibility of split band InSAR processing using TerraSAR-X Spotlight data as also the importance of the information content of spectral coherence within a single acquisition [9, 10]. Next, in the frame of the Vi-X project (BelSPo contract N° SR/00/150), SBInSAR technique was also demonstrated using TanDEM-X data. In the latter case, it was shown the importance of controlling the co-registration process [11]. The lack of coherence due to time de-correlation in the first case or the imprecision due to the too short available bandwidth in the second one prevent reaching the required accuracy on the estimated absolute phase.
In the present project, we propose to go the required step further using wider bands and/or highly coherent data to reach the absolute phase measurement on available seed points and integrate these seeds point in a classical phase unwrapping process to generate continuous absolute phase maps, connecting isolated areas due to local coherence loss or geometrical deformation preventing phase integration, for either topographic or displacement measurements.
OBJECTIVES
The main objective of the MUZUBI project is to make the required step forward, taking advantages of the results obtained during the Vi-X project, to implement absolute phase unwrapping wherever possible. The proposed method will furthermore improve the selection of pixels that remain coherent over time, adding spectral stability, which is an important aspect for the time series methods such as Small Baselines Subset (SBAS) [1, 2].
This MUZUBI project is linked to an accepted project (ViCoX) submitted in the frame of the TanDEM-X science phase AO, granting a free of charge delivery of an exceptional dataset made of stripmap and spotlight pursuit mode acquisitions (project ID: NTI_INSA6714) Split-Band SAR interferometry (SBInSAR) is based on spectral analysis performed on each image of an InSAR pair, yielding a stack of sub-band interferograms. Scatterers keeping a coherent behaviour in each sub-band interferogram show a phase that varies linearly with the carrier frequency, the slope being proportional to the absolute optical path difference. If the available range spectral bandwidth is large enough and if spectral an interferometric coherence are sufficiently preserved, linear regression of the phase ramp through the stack may lead to an estimation of the absolute interferometric phase.
The Vi-X project definitively showed the potential of Split Band InSAR processing to derive absolute phase measurements. It was shown during this project that SBInSAR processing is first a drastic oregistration improvement of interferometric SAR images, which implies having a thorough knowledge of the applied co-registration within the InSAR processing. This co-registration improvement may go up to the absolute phase estimation provided the signal bandwidth allows performing an accurate spectral analysis of the signal.
The main objective here is to find such points allowing performing this absolute phase estimation, almost up to the correct fringe index. Classical interferometric phase will give the fractional part of the absolute phase. These points will then be used as seeds points within the classical phase unwrapping process to extend by spatial integration the absolute phase and get a continuous phase surface.
Therefore, from a processing and implementation point of view, the main objective will be the integration of both techniques to get an interferometric processing chain allowing to get an absolute phase map wherever possible.
Intermediates objectives will be:
– To adapt existing SBInSAR processing to the specific case of TerraSAR-X pursuit mode and spotlight mode.
– To determine what are the required signal bandwidth and splitting scheme with respect to local spectral and temporal coherence to find seeds points with the required accuracy; it is to say a phase estimate with accuracy better than half a fringe to almost find the corresponding fringe index.
– To integrate SBInSAR processing within the CSL InSAR suite and merge it with classical phase unwrapping processing.
– To test the technique to derive either a displacement map (Nyiragongo study case) or a topographic phase component (Copahue study case).
– To validate derived displacements in the case of the Nyiragongo volcano (link with the RESIST project “Remote Sensing and In Situ detection and Tracking of geohazards”) by comparing derived displacement map with those resulting from Multidimensional Small Baseline Subset (MSBAS) [3] processing chain developed at ECGS/MNHN Luxemburg.
– To validate the derived topographic measurement in the case of the Copahue volcano.
METHODOLOGY
This spin-off project is basically a follow-up of the Vi-X project (BelSPo contract N° SR/00/150), which aimed at developing and adapting Split Band Interferometry (SBInSAR) to the TanDEM-X bistatic case.
Thematic objective consisted in monitoring the rapid topographic changes associated to the fast (though un-monitored) ~50m/yr rising of the lava-lake located inside the Nyiragongo crater. During this mother project, SBInSAR processing was fully implemented, including TanDEM-X coregistration data handling. It was shown that SBInSAR applied to already co-registered data allowed obtaining only one component of the absolute phase. The first phase component, called the registration phase, must be computed knowing exactly the registration shift applied to the salve image with respect to the master one. Adding the residual phase component derived through SBInSAR allows getting the full phase on a point-by-point basis.
However, it appeared that in the case of available images of the Nyiragongo volcano, the spectral diversity resulting from the bandwidth (100MHz) of TanDEM-X does not allow getting absolute phase accuracy better than about 3 fringes in the most favourable cases.
Consequently, if the potentiality of SBInSAR was clearly demonstrated, in the case of the used data set, the available bandwidths still not allow getting the required phase accuracy to perform a point-wise absolute phase unwrapping to connect independently-unwrapped zones, even if the local coherence is high.
It is expected that one can reach the required precision using either TanDEM-X data in pursuit mode or in bistatic spotlight high-resolution mode, which is the reason why a project named ViCoX: “Study and monitoring of Virunga and Copahue volcanoes using TanDEM-X” was submitted to DLR in response to the recently launched “TanDEM-X Science Phase Announcement of Opportunity”. This submitted ViCoX project was accepted, which will grant us access to exceptional data sets in stripmap and spotlight pursuit mode. These acquisition modes offer a larger range bandwidth (from 150Mhz up to 300MHz) that will allow improving the absolute phase measurement.
Since absolute phase ranging principle was clearly demonstrated, it is proposed to fully implement and integrated it as a precursory step in the phase unwrapping procedure in order to get a continuous phase surface wherever possible. Next, this unique developed tool will be used on two different test sites using TanDEM-X science phase datasets.
On one hand, Nyiragongo test site is highly vegetated and offers a reasonable topography except for its crater and lava lake pit. On the other hand, the Copahue area shows a highly dynamic relief in a semi arid and geothermal environment. In the first case, it is expected that, despite the potential losses of temporal and geometrical coherence due to volumetric scattering within the vegetation, the stripmap and spotlight pursuit mode data will allow reaching the required precision to perform a valuable lava lake levelling. In the second case, it is expected that these data will allow performing a complete DEM of the area.
WP0 – Management
WP1 – Adapting SBinSAR method: First step will consist in adapting the existing SBInSAR processing to TanDEM-X stripmap and spotlight pursuit mode data specificities. Results from previous Vi-X project demonstrated the importance of mastering the applied co-registration. Therefore, if pursuit mode data are provided already co-registered, we will implement the required routines to handle it and generate the registration phase. This approach might be extremely important in the future if willing to use Sentinel-1 data that are potentially good candidate data for SBInSAR processes (bandwidth over carrier frequency ratio is equivalent to classical CSK and TSX data).
WP2 – Assisting unwrapping using SBInSAR: With respect to integration of SBInSAR processing within the phase unwrapping process, the methodology will be to detect seed points on which SBInSAR processing is performed with the required accuracy. These seed points must have a phase standard deviation on the intercept of the linear regression of less than a fringe to grant finding almost the fringe index to which belongs the considered point. The fractional part of the phase will be given through classical interferometry. Classical phase unwrapping based on residue connection and spatial integration will lead to patches of independent and unconnected unwrapped areas. Within a given independent zone, the difference between the absolute phase of each seed point and its classically-unwrapped phase should lead to a constant integer number of fringe corresponding to the phase shift to be applied to the considered zone. If the number of seed points within a zone allows it, a histogram of this phase shift will be drawn to find the most probable one. This histogram will allow refining the required phase standard deviation to really find the phase shift to be applied. Seed point listing for the whole scene will be refined to cope with this corrected phase standard deviation threshold. Consequently, starting from the biggest to the smallest zone, seed points listing will be refined to use only those having really the required accuracy to find the phase shift to be applied to each zone in order to get a corrected phase surface.
The developed and integrated SBInSAR improved phase unwrapping processing chain will be used on specific TanDEM-X and TerraSAR-X Scientific data sets to generate the absolute phase map on two test sites: Nyiragongo & Copahue.
WP3 – Application to Nyiragongo test site: Application of the implemented technique on the data sets available on the first test site will be done in two steps. The first step will be based on stripmap and spothlight pursuit mode data and will consist in finding the best parameters to optimize the number and the accuracy of found seed points. SBInSAR main parameters are sub-band bandwidth and number of sub-bands. The first one determines the resolution cell of the final product and the width of the explored bandwidth. The second one determines the number of sample points in the spectral domain to perform linear regression on the observed phase trend. This second factor is determinant in terms of computational load. This step will allow optimizing these parameters for each acquisition mode.
Second part will consist in using found parameters to issue pursuit mode topographic phase and perform lava lake levelling, measuring the relative height difference between points inside of the crater on bottom platform P3 and some other known area in the surrounding of the volcano.
That application will finally involve field observations dedicated to improve the interpretation of the results and explore possible validations processes. External methods such as DGPS measurements for topography, close range photogrammetry for studying the lava lake activity and associated phenomena.
WP4 – Application to Copahue test site: Using the optimized parameters found on WP3, test will be performed on Copahue test site to verify that parameters are acquisition mode dependent and not scene dependent. Produced topographic phase and related DEM from pursuit mode data will be validated comparing it with respect to an external DEM (SRTM). Even if not as precise in terms of resolution cell and height accuracy, it will allow detecting rough errors in phase shifts applied to independent zones. In a second step, these zones will be verified in priority through ground truth measurements.
Second part of WP4 will consist in fine validation of derived topographic information with ground truth measurements.
WP5 – SBInSAR applicability in Differential Interferometry context: SBInSAR processing will be performed on data taken some cycles apart to detect possible displacements in the Nyiragongo-Nyamulagira volcanic field, if any, and test the applicability of SBInSAR-improved phase unwrapping to connect independent areas of displacement phase in lower coherence context due to unavoidable time decorrelation. In connection with the RESIST project, SBInSAR-assisted unwrapped phase maps will be compared with results obtained from the Multidimensional Small baseline Subset (MSBAS) InSAR processing chain developed at ECGS/MNHN.
Consistancy between both techniques will be assessed. Although not critical, complementary ground truth performed in the frame of mother project could also provide additional constrain to this crossvalidation process.
WP6 – Dissemination and valorization toward end-user: Dissemination will be granted through the ViCoX project submitted to DLR. Results will be presented in this frame during a scientific meeting organized by DLR. In addition, results shall also be presented in ESA FRINGE symposium. Almost one publication per year shall be submitted to a peer reviewed publication. A specific web page linked to the RESIST one will be maintained.
If Nyiragongo lava lake levelling appears to be successful thanks the MUZUBI outcomes, a large audience diffusion of the results shall be done, in collaboration with the RESIST project.
EXPECTED SCIENTIFIC RESULTS
The main outcome is the development of an integrated absolute phase unwrapping processing. It is expected to develop SBInSAR and to integrate it in existing processing chain to allow getting continuous phase measurement, resolving phase ambiguities between isolated or independent areas.
Using TerraSAR-X Science phase date (i.e. spotlight and pursuit mode), it is expected to generate a resolved displacement maps of the Nyiragongo volcano. These results will then be validated and integrated into the RESIST project to improve the analysis of spatio-temporal distribution of ground deformation associated to volcanic activity.
In the case of the Copahue volcano, it is expected to generate and validate a high-resolution topographic map of the whole area, resolving phase ambiguities in this difficult area due to geometrical deformations. In turns, this topographic map will then be used by LGA in their water management models.
EXPECTED PRODUCTS AND SERVICES
First expected product is a method allowing to solve or ease the phase unwrapping process, extending the measurement area, whatever the final use of the derived phase surface (topographical measurements or displacement monitoring).
Based on the developed method, derived products will be:
• A better lava lake levelling of the Nyiaragongo volcano to improve its monitoring.
• An extended DEM of the Copahue volcano
Since lava lake levelling requires a constant monitoring, the developed methodology may also be considered as a potential service
References
[1] J.C. Curlander, R.N. McDonnough, “Synthetic aperture Radar”, Wiley Interscience, New York, 1991.
[2] A. Arnaud, “Ship detection by sar interferometry” in IEEE 1999 Inter- national Geoscience and Remote Sensing Symposium, 1999. IGARSS’99 Proceedings, vol. 5, 1999.
[3] N. Bechor, and H. A. Zebker, “Measuring two-dimensional movements using a single InSAR pair”, Geophys. Res. Lett., 33, n°16, 2006, doi:10.1029/2006GL026883.
[4] Cumming, I., Wong, F., & Hawkins, B., ‘‘RADARSAT-1 Doppler Centroid Estimation Using Phase-Based Estimators’’, SAR Workshop: CEOS Committee on Earth Observation Satellites; Working Group on Calibration and Validation, 26-29 Oct. 1999. ESA-SP vol. 450, ISBN: 9290926414, p.159
[5] S. N. Madsen and H. A. Zebker, ‘‘Automated Absolute Phase Retrieval in Across- Track Interferometry’’ Proc. IGARSS’92, vol. II, pp. 1582–1584, 1992.
[6] N. Veneziani, F. Bovenga, A. Refice, “A wide-band approach to absolute phase retrieval in SAR interferometry”, Multidimensional Systems and Signal Processing, MULT vol. 14(1-2), pp. 183-205, Kluwer Academic Publisher, 2003.
[7] R. Bamler, M. Eineder, “Split Band Interferometry versus Absolute Ranging with Wideband SAR Systems”, Proc. IGARSS-04, Anchorage (Alaska), September 2004.
[8] G. Gomba; M. Eineder, “Correction of Ionospheric Phase Screen in Differential Interferograms Using Range Split-Band Technique”, Proc. FRINGE 2015, March 2015, ESA SP-731
[9] D. De Rauw, A. Orban, Chr. Barbier, “Wide band SAR sub-band splitting and inter-band coherence measurements”, Remote Sens. Letters, Vol. 1, 3, pp 133-140, Sept. 2010 (http://hdl.handle.net/2268/1880)
[10] F. Bovenga, Chr. Barbier, D. De Rauw, F. Rana, A. Refice, N. Veneziani, R. Vitulli, “Multi-chromatic Analysis of SAR Images for Coherent Target Detection”, Remote Sens., vol. 6, pp 8822-8843, September 2014 (http://hdl.handle.net/2268/173612)
[11] D. Derauw, F. Kervyn, N. d’Oreye, B. Smets, F. Albino, Ch. Barbier, “Split-Band Interferometric SAR Processing Using TanDEM-X Data’, Proc. FRINGE 2015, March 2015, ESA SP-731 (http://hdl.handle.net/2268/179826)