Besides the expertise pools at IMF in the three sensor technologies, namely synthetic aperture radar (SAR), optical imaging and atmospheric spectrometry, algorithms crossing sensor technologies, e.g. for data fusion, have been gaining importance for years. Furthermore, Earth observation (EO) has irreversibly arrived in the Big Data era with the Sentinel satellites (and in the future with Tandem-L). This requires not only new technological approaches to manage large amounts of data (as pursued by DFD), but also new analysis methods.
Here, methods of data science and artificial intelligence (AI), such as machine learning, become indispensable. Deep Learning in particular has led to a revolution in AI in recent years. Motivated by these facts, IMF has placed one of its research foci on EO Data Science.
Our vision is listed as follows:
Most of the operational data processors for EO missions are model-based algorithms. We have been developing algorithms in sparse reconstruction in SAR interferometry and optical images, robust estimation, nonlocal image filtering, and tensor analysis over the past decade. We will continue develop such modern signal processing algorithms in the future.
We focus on the exploration of AI for EO
Our current focus is particularly on AI for multi-modal data, geo-reference remote sensing data, image time series, and large-scale data.
We develop sophisticated algorithms and discover novel applications crossing sensor technologies
This includes the fusion of SAR and optical images, the fusion of InSAR and optical 3D point clouds, as well as the fusion of polarimetric SAR, multi-/hyper-spectral image, and LiDAR for classification tasks, and many more.
We have decades of experience in data mining and knowledge discovery of remote sensing data. In the future, we will pursue a full ecosystem from algorithm development, to high performance computing, and eventually to geoscientific applications. This requires the implementation of high performance codes, modern remote sensing data structure such as the Datacube, and user-friendly platforms.
These unconventional data include but not limited to street level image, e.g. Google street view image, social media text messages, and micro/nano satellite images, e.g. Planet data.
The department consists of two teams
In addition, the professorship of Signal Processing in Earth Observation (SiPEO) in Technical University of Munich, which is held by the department head, also shares the same abovementioned vision. For details see Link.