
High-resolution satellite imagery has become essential for providing detailed observations of terrestrial surfaces over time. Frequent revisits to the same geographical areas, made possible by satellite constellations such as Sentinel and Landsat, generate spatio-temporal data known as Satellite Image Time Series (SITS), which contain valuable temporal patterns. However, analyzing SITS presents significant challenges due to their large data volume, complexity, and variability. This talk will start with an introduction to the SITS research field, highlighting key advances and current perspectives. From there, we will delve into three main topics. First, we will discuss advanced approaches for SITS classification based on graph and covariance matrix representations. Next, we will present a novel super-resolution strategy that enables the generation of high spatial resolution images from lower-resolution SITS. Finally, we will cover two domain adaptation techniques specifically tailored for SITS, which allow extending predictions to new geographical areas where labeled data is unavailable for training supervised models.

Bio:
Charlotte Pelletier is an associate professor in computer science at Univ. Bretagne Sud in Vannes, France, with a PhD from Univ. Toulouse III, specializing in the analysis of satellite image time series. Her research at the Institute of Research in Information Technology and Random Systems (IRISA) focuses on developing new accurate and scalable time-series classification algorithms, time-series domain adaptation and super-resolution applied to Earth observation data. She is coordinating a four-year project (2024-2027) on continual learning for satellite image time series funded by the French Research Agency (ANR) and is involved in various French and European projects. Charlotte has chaired a working group on temporal geospatial data understanding (ISPRS TCII/WG5) since 2022 and co-chaired a technical committee on remote sensing and mapping (IAPR TC7) since 2019. She has been involved in the organization of several special sessions and workshops at international venues, including Earth Vision (CVPR) since 2020.