Exploiting sparsity in Remote sensing and Earth Observation

Speaker: Dr. Xiaoxiang Zhu (DLR)

Time and Date: 10:00-12:00, Jan. 14, 2014

Place: Room 1101, East Guanghua Building, Handan Campus

 

 

Abstract

Remote sensing enables us to recover contact-free large-scale information of objects on the earth. It can be widely employed for civil engineering and geoscientific applications, such as city planning, urban infrastructure monitoring, traffic monitoring, mass variations of glaciers, natural disaster prevention, geo-information systems.

 

Many modern advanced applications in remote sensing call for sensors with ever high resolution. This would lead to excessively high requirements on the involved hardware and software due to the huge amount of data sampled according to Shannon’s sampling theorem. However, this paradigm neglects the fact that many signals of interest are sparse, i.e. although not necessarily band-limited but can still be represented by a model with a small number of degrees of freedom. Sparse signals are commonly expected in earth observation. E.g. radar images have much less information content than the acquired raw data samples pretend. Another example is hyperspectral unmixing where only few materials are expected in a pixel compared to the prodigious endmember library. Sparse reconstruction utilizes this sparsity property and allows reconstruction of signals from much fewer samples than the conventional sampling theory requires or it can give higher resolution than the Rayleigh limit suggests. We develop sparse reconstruction based algorithms for several problems that cover a wide range of fields including SAR and optical (multispectral and hyperspectral) remote sensing. The developed algorithms are evaluated with both simulated and real remote sensing data, e.g. acquired by spaceborne systems such as TerraSAR-X, TanDEM-X, and WorldView-2 and by airborne sensors such as HyMap and HySpex.

 

 

Biography

Xiaoxiang Zhu received the Master (M.Sc.) degree, her doctor of engineering (Dr.-Ing.) degree and Habilitation from Technical University of Munich (TUM), Germany, in 2008, 2011 and 2013, respectively. Since May 2011, she is a scientist with the Remote Sensing Technology Institute at the German Aerospace Center (DLR), Oberpfaffenhofen, where she is the head of the Team Signal Analysis and with the Chair of Remote Sensing Technology at TUM. Since September 2013, she is also a Helmholtz Young Investigator Group Leader. Her main research interests are: advanced InSAR techniques such as high dimensional tomographic SAR imaging and SqueeSAR; computer vision in remote sensing including object reconstruction and multi-dimensional data visualization; and modern signal processing, including innovative algorithms such as compressive sensing and sparse reconstruction, with applications in the field of remote sensing such as multi/hyperspectral image analysis. She is giving courses in signal processing, estimation theory and remote sensing. Dr. Zhu is the developer of the DLR's Tomographic SAR Processing System ─ Tomo-GENESIS. She is author of more than 80 scientific publications, among them 39 full-paper peer-reviewed papers and 6 paper awards.