Skip to Main content Skip to Navigation

Reconstruction de données manquantes dans des séries temporelles de mesures de déplacement par télédétection

Abstract : Despite the large volume of available data (satellite and in-situ) in displacement measurement, data incompleteness is still a commonly encountered issue. This phenomenon is mainly due to surface property changes of the observed object and/or to technical limitations of the displacement extraction methods (e.g. differential interferometry, offset tracking). By generating time and space discontinuity, data incompleteness can hinder a thorough understanding of underlying physical phenomena that induce surface displacement. However, missing data analysis in displacement measurement has not been paid significant and dedicated attention. In this context, advanced reconstruction methods, adapted to the data specificities, are necessary for handling spatio-temporal gaps in displacement measurement time series. In this thesis, we propose three approaches for analysing and imputing missing data in remotely sensed displacement measurement time series. The two firstapproaches are based on the decomposition of the temporal and spatio-temporal covariance of the displacement signal into Empirical Orthogonal Functions (EOFs). These studies have led to the development of two methods, so-called EM-EOF and extended EM-EOF, both requiring an initialization of the missing values. The third approachintends to explore techniques in robust estimation of the covariance matrix without initialization of the missing values. All approaches rely on an Expectation-Maximization (EM)-type iterative resolution scheme and reckon with the covariance low rank structure, which describe most of the variability of the displacement signal. Both synthetic simulations and real data applications present promising results, bringing to light the interest of the proposed approaches for missing data imputation in remotely sensed displacement measurement time series
Complete list of metadata
Contributor : ABES STAR :  Contact
Submitted on : Monday, January 3, 2022 - 12:32:12 PM
Last modification on : Wednesday, March 2, 2022 - 3:41:06 AM
Long-term archiving on: : Monday, April 4, 2022 - 7:01:25 PM


Version validated by the jury (STAR)


  • HAL Id : tel-03507672, version 1



Alexandre Hippert Ferrer. Reconstruction de données manquantes dans des séries temporelles de mesures de déplacement par télédétection. Algorithme et structure de données [cs.DS]. Université Savoie Mont Blanc, 2020. Français. ⟨NNT : 2020CHAMA027⟩. ⟨tel-03507672⟩



Record views


Files downloads