ESA Earth Home Missions Data Products Resources Applications
EO Data Access
How to Apply
How to Access
Site Map
Frequently asked questions
Terms of use
Contact us



Optimal data generation for Multiple Images DInSAR techniques: Minimum Cost Flow interferograms and Region Growing pixel selection.

Pablo Blanco-Sanchez(1), Sergio Duque(1), Jordi J. Mallorqui(1) and Dani Monells(1)

(1) Universitat Politecnica de Catalunya, Campus Nord UPC D3, 08034 Barcelona, Spain


Differential interferometry has become one of the most important applications of SAR (Synthetic Aperture Radar) data. In fact, Multiple-Images DInSAR algorithms, developed during this past decade, have proven to be reliable techniques capable of monitoring deformation phenomena very precisely. The main product of these techniques is the generation of deformation and subsidence maps of the observed area. Terrain deformation is valuable information for risk management and for the understanding of much geological process.

Prior to extract useful information from the interferometric phase, it is mandatory to handle the available image dataset and transform it into meaningful data. This manipulation can be addressed in two steps: creating the set of interferograms to work with and selecting the quality pixels among the chosen stack of interferograms. In this paper we will tackle both topics, proposing a Minimum Cost Flow (MCF) based method for selecting a reduced and quality set of interferograms, and extending a Region Growing (RG) based pixel selection criterion to the Multiple Images DInSAR framework. Furthermore, these topics will be tested with the Coherent Pixels Technique (CPT) developed at UPC ([1],[2],[3]).

1) Interferogram set Selection.

As the historical archive become larger (close to 15 years in ESA’s SAR missions) the amount of SAR data available nowadays from a particular scenario can be really huge. Reducing this large number is important in order to perform a quick (with low disk storage requirements) processing in order to identify a problematic area and give a first characterization of the deformation phenomena going on. Consequently, we wanted to investigate the further we could go with CPT dealing with a reduced set of interferograms for generating reliable results and the best way to generate this dataset. Evidently, once the problematic area is detected, large datasets are welcome to deeply analyze the selected scenario.

CPT selects the interferogram set to work with by performing a 3D Delaunay triangulation of the available images in the {T, Bn, fdc} space, so the number of possible interferograms is considerably reduced while their coherences are maximized. In order to even reduce this number of interferograms a second selection can be performed among the arcs (interferograms) of the triangulated mesh. The goal is to obtain the best set of interferograms that defines the minimum number of equations to solve the deformation problem and cancelling redundant information (this is N-1 linearly independent interferograms form a dataset of N images). Here, a Minimum Cost Flow (MCF) based algorithm has been applied, where we have also taken care of ensuring a uniform {T, Bn, fdc} final distribution. The theoretical coherence calculated in function of the spatial and temporal baselines and the Doppler centroid difference for each interferogram is chosen as the cost function, showing agreement with the highest number of top quality pixels per interferogram, as validation with real data tell us so.

2) Pixel Selection Criterion.

When dealing with a reduced number of interferograms (especially in that case, but not only), it is also be desirable to have a pixel selection method capable of accurately discriminate high quality from low quality pixels within the selected interferograms. As an alternative to the classical pixel selection method based on the coherence stability and the dispersion of amplitude, a Region Growing (RG) [4] based method has been extended to the Multiple Image DInSAR framework. This method has the ability to search those pixels which present a similar statistical behaviour to the initial one, avoiding constant multi-look factors.

Restricting this search within the same multi-looking window, two very interesting properties come out to light. First, spatial resolution is increased, as the resultant phase is averaged in a minor number of pixels, and second, the resultant phase is less noisy, as averaging is performed only by statistically similar pixels. The performed tests have shown great performances for quality deformation movement retrieval.

The results to be presented have been obtained from ERS and ENVISAT data provided by ESA over different test-sites (Paris, Gardanne, Murcia) in the scope of different projects, as the DInSAR algorithms intercomparison PSIC4 promoted by ESA.

[1] Mora, J.J. Mallorquí, T. Broquetas, “Linear and nonlinear terrain deformation maps from a reduced set of interferometric SAR images”, IEEE Trans. Geosci. Remote Sensing, Vol. 41, 2243-2253, 2003. [2] P. Blanco, J.J. Mallorquí, D. Navarrete, S. Duque, J. Sanz-Marcos, P. Prats, R. Romero, J. Dominguez, D. Carrasco, “Application of the Coherent Pixels Technique to the Generation of Deformation Maps with ERS and ENVISAT Data”, International Geoscience and Remote Sensing Symposium, 2005. IGARSS '05. Vol 3, pp:1983 – 1986. [3] P. Blanco, J.J. Mallorquí, S. Duque and D. Navarrete, “Advances on DInSAR with ERS and ENVISAT data using the Coherent Pixels Technique (CPT)”, International Geoscience and Remote Sensing Symposium, 2006. IGARSS '06.

[4] G. Vasile, E. Trouvé, J.S. Lee and V. Buzuloiu, “Intensity-Driven Adaptative-Neighborhood Technique for Polarimetric and Interferometric SAR Parameters Estimation”, IEEE Trans. Geosci. Remote Sensing, Vol. 44, 1609-1621, 2006.


Workshop presentation

Keywords: ESA European Space Agency - Agence spatiale europeenne, observation de la terre, earth observation, satellite remote sensing, teledetection, geophysique, altimetrie, radar, chimique atmospherique, geophysics, altimetry, radar, atmospheric chemistry