Showing papers by "Florence Tupin published in 2005"

Markov random field on region adjacency graph for the fusion of SAR and optical data in radargrammetric applications

Abstract: This paper deals with the estimation of an elevation model using a pair of synthetic aperture radar (SAR) images and an optical image in semiurban areas. The proposed method is based on a Markovian regularization of an elevation field defined on a region adjacency graph (RAG). This RAG is obtained by oversegmenting the optical image. The support for elevation hypotheses is given by the structural matching of features extracted from both SAR images. The regularization model takes into account discontinuities of buildings thanks to an implicit edge process. Starting from a good initialization, optimization is obtained through an iterated conditional mode algorithm.

An empirical model for interferometric coherence

Abstract: Many are the examples of application of SAR and differential SAR interferometry for topographic mapping and ground deformation monitoring. However, on repeat pass geometry, the performances of these two techniques are limited by the loss of correlation (coherence) between the two radar acquisitions. The lack of coherence causes an additional noise thus a poor estimate of the interferometric phase. The disturbances can be due either to surface changes because of long period cover (temporal decorrelation) or to a too large baseline (spatial decorrelation). In this paper, we propose an empirical model for the estimate of coherence considering separately these two sources of disturbances. Starting from the observations of experimental data, we study the behaviour of coherence according to baseline and period cover in order to express the two terms of correlation. A number of 170 multi temporal and multi baseline differential interferograms covering the same region is used to validate the proposed model.

Registering of synthetic aperture radar and optical data

Abstract: The registration of Synthetic Aperture Radar (SAR) and optical images is a challenging problem, notably in the perspective of the future launch of the Cosmo-SkyMed sensors in the framework of Orfeo program, which will provide high resolution images (below 1 meter). In this paper, we first present the registration of SAR and optical images using a perfect knowledge of the sensor parameters (position, speed, etc, but also the ground height) and we study the influence of a bad knowledge of these parameters. One then shows that feature extraction can be used to perform parameter refining. Finally, one points out that this SAR and optical image registration can be approximated by simple polynomial transformations.

Validation of a feature fusion scheme for urban DSM retrieval from high resolution SAR interferogram

Abstract: Three-dimensional reconstruction in urban areas is one of the major issues in remote sensing applications. SAR interferometry (InSAR) has a great potential to provide Digital Surface Models (DSM), but the context of urban areas and high resolution (HR) is difficult (geometrical distortions, surface heterogeneities, speckle, height discontinuities). Previous studies prove both the difficulty and the potential of the method. In this paper, we propose a new fusion method to compute DSM from HR InSAR. The merging approach enables to take into account several kinds of information extracted from the original data in order to retrieve jointly a classification and a DSM. The paper will be focused on the analysis of the results in order to understand the limits and potentials of InSAR over urban areas. Before discussing the results, the method will be briefly described.

A Markovian scheme for joint retrieval of classification and height map from urban interferometric SAR images

Abstract: Synthetic aperture radar (SAR) interferometry enables to compute an height map of the scene which is useful for many applications. Yet the complexity of high resolution SAR images and of urban areas prevents from computing an accurate DSM easily and a high level processing chain is thus required. In this article, we propose a Markovian fusion scheme to retrieve jointly the height map and the classification. The original data (amplitude, interferogram and coherence) are first processed in order to get new entries. They represent advanced information on the scene, extracted with different approaches (filtering, object recognition or global classification). These features are then merged in a Markovian framework to recover an improved classification and height map. The method is illustrated on real data.