Deep learning for Satellite Image Time Series Analysis
Synopsys: Dynamics on the Earth’s surface are governed by continuous temporal processes that can be observed in discrete intervals by Earth observation satellites. Recent satellite constellations, such as Landsat-8, Sentinel-1 and 2, produce a high volume of satellite image time series by covering the same location on Earth at frequent temporal intervals. The increase in the number of acquired images, the diversity of the time series (e.g., optical and radar), and the combinations of the high spatial and high temporal resolutions enable advances in a variety of applications, such as vegetation modeling, climate forecasting, urban planning, or precipitation nowcasting. The complexity of the data (irregular temporal sampling, multi-modality, multispectral data, high volume of data, low number of high-quality reference data) requires the development of novel data-driven methods to solve (early) classification, regression, forecasting, or indexing tasks. In this thematic session, we welcome new contributions that advance the analysis of satellite image time series.
The session is open to all authors.