Remote sensing is today crucial for many domains including environmental monitoring, natural disaster management, agriculture, urban planning, safety and security.  The emergence of novel satellite constellations, rapid progress with unmanned area vehicles (UAV) and acquisition of novel types of imaging modalities with increasing spatial resolution and temporal revisit frequencies brings unprecedented potentials for various aspects of Earth observation, including climate change monitoring. At the same time, we experience incredible challenges for data processing and extraction of the information of interest from these vast amounts of heterogeneous data.

The overall goal of REALISE is to develop a solid theoretical framework for reliable unsupervised change detection from multi-sensor, multi-temporal remote sensing data. At present, no solid methodology exists yet for change detection from multiple multi-source images, let alone dealing with irregularly sampled and partially missing modalities of different types. This project thus has the potential for important breakthroughs both regarding the fundamental techniques and applications in Earth observation.