Science Service System

Summary of Proposal MTH1595

TitleSegmentation and Classification of HS TerraSAR-X data
Investigator Nies, Holger - University of Siegen, Center for Sensorsystems (ZESS)
Team Member
Prof. Dr.-Ing. Loffeld, Otmar - University of Siegen, Center for Sensorsystems (ZESS)
M. Sc. Yao, Wei - University of Siegen, Center for Sensorsystems (ZESS)
Prof. Dr.-Ing. Datcu, Mihai - German Aerospace Center (DLR), Remote Sensing Technology Institute
SummaryThe main issue of our project is to build up an algorithm to perform land cover classification task over urban area. We have started to develop a classification algorithm which aims at high resolution SAR images. As the TerraSAR-X satellite already generated a lot of high resolution products, we intend to utilize these high resolution products to train and test our classification algorithm, within the scenes of the whole of Germany and some interesting parts of Europe. The algorithm framework will not only choose and combine suitable learning algorithms (e.g. SVM, Adaboost, topic model, etc.), but also select useful features (e.g. textual, SIFT, etc.).
Since in our institute, we have already generated several bistatic products, we can also test our algorithm to see whether it can be adapted to the classification of bistatic products. We plan to make a limited comparison in those scenes where we have both bistatic and monostatic products. In the final step we plan to build a machine learning system which is capable of individually handling both monostatic and bistatic data. Moreover, we can try to extend the algorithm to integrate both sources.

Back to list of proposals

DLR 2004-2016