Traitements cohérents d’images RSO multi-modes multi-resolutions pour la caractérisation du milieu urbain

Abstract : The refinement of Synthetic Aperture Radar (SAR) imagery makes the monitoring of urban areas possible. Different acquisition modes of multivariate SAR images are available, such as the polarimetric mode that gives information about the nature or the geometry of imaged objects, or the interferometric mode that enables the measure of their height or the displa- cement between two acquisitions. However, all the acquisition modes do not reach the same spatial resolution. The major goal of this work was to combine polarimetric images and a mono-channel high-resolution image to create high resolution polarimetric images with the same properties as the initial polarimetric image. This work started with the study of covariance matrices estimation from which meaningful parameters linkable to physical infor- mation, such as the phase, can be extracted. We first showed that the variance is not adapted to describe the fluctuations of the estimated phase. We proposed a directional statistics indicator to describe the estimated phase fluctuations, whichever the definition interval of the phase. This indicator can be used to separate two populations of pixels with different phase fluctuations or to compare the measured phase to a modeled phase. We also showed that the resolution has an impact on the covariance matrix estimation, because it restrains the size of the sample of pixels which values can be considered as independent and identically distributed for the covariance matrix estimation. We chose to model the urban scatterers as a combination of randomly responding scatterers and deterministic bright point-like scatterers. The values of the pixels representing the random scatterers are modeled following the fully developed speckle ; the point-like scatterers are supposed to be white and isotropic. The Pol-Sharpening algorithms that we developed are based on the point-like scatterers spectral properties. One of the key issue is to have a sufficient sample size in order to limit the fluctuation of the polarimetric information estimation without introducing mixing. The chosen spectral method enables to focalize the point- like scatterers, i.e. to refine the resolution, while preserving their polarimetric and interferometric properties. If the developed algorithms keep the value of intensity and phase of the pixels included in homogeneous speckle areas, they modify their statistical parameters. Finally, we studied the measure of the height and the deformation of the Eiffel Tower and three other towers on the Front de Seine in Paris using a temporal stack of 98 TerraSAR-X images acquired between 2008 and 2012. We showed that neglecting the deformation during the height measurement could result in outliers. Even though these outliers can be removed during the study of a temporal stack, it is still a limitation of the interferometry techniques on buildings. By measuring a monoscopic height, we could measure deformations up to four centimeters for the Eiffel Tower. These deformations do not present any seasonal trend contrary to the deformations measured on the Mirabeau Tower, which are minimal in winter and maximal in summer. Nevertheless, we neglected the impact of the atmosphere on the deformation measurement and this work hypothesis limits our measurement accuracy to the centimeter.
Liste complète des métadonnées
Contributor : Flora Weissgerber <>
Submitted on : Thursday, March 2, 2017 - 4:51:48 PM
Last modification on : Tuesday, March 26, 2019 - 2:24:45 PM


  • HAL Id : tel-01481528, version 1



Flora Weissgerber. Traitements cohérents d’images RSO multi-modes multi-resolutions pour la caractérisation du milieu urbain. Traitement du signal et de l'image. Télécom ParisTech, 2016. Français. ⟨NNT : 2016ENST0074⟩. ⟨tel-01481528⟩



Record views


Files downloads