Title: Iterative robust hypothesis testing for change-points detection and application to SAR change detection
Authors: Ammar Mian - CentraleSupelec SONDRA (France) [presenting]
Guillaume Ginolhac - LISTIC - Universite Savoie Mont-Blanc (France)
Jean-Philippe Ovarlez - ONERA Palaiseau and CentraleSupelec SONDRA (France)
Abdourahmane Atto - LISTIC - Universite Savoie Mont-Blanc (France)
Abstract: The problem of detecting change-points in a time series of multivariate sets of vectors is considered by exploiting covariance homogeneity testing schemes. Notably, we propose to consider the problem under a robust framework by assuming Complex Elliptically Symmetric distribution models for which we tailor detection tests. The robustness of this approach, compared to one based on a Gaussian assumption, will be considered through theoretical and experimental analysis. Then, based on these developments, we consider an iterative algorithm to determine the points of change in a time series of data. Finally, an application to the analysis of changes in Synthetic Aperture Radar time series will be proposed to demonstrate the interest of the robust framework in real-world applications.