Science Service System
You are here : Home : Proposals_Summary

Summary of Proposal MTH0288

TitleAutomatic Object Categorization in High Resolution SAR Data
Investigator Kaufmann, Hermann - GeoForschungsZentrum Potsdam, Geodesy and Remote Sensing
Team Member
Prof. Dr. Hellwich, Olaf - Berlin University of Technology, Computer Vision & Remote Sensing
SummaryThe project is concerned with the development of object categorization methods for high resolution SAR and polarimetric SAR data. Recent years have seen remarkable advances in SAR technology (Terra SAR X being the prime example). Simultaneously, the computer vision community has developed generic techniques for categorizing objects in images based on object models trained without supervision. The aim of this project is the application of such techniques in the automatic analysis of high resolution SAR data containing highly complex structures. This constitutes the first integration of two lines of research: on the one hand, computer vision approaches based on machine learning techniques have, to date, not been applied in the analysis of SAR data, and on the other hand modern image analysis tools have rarely been applied in the remote sensing context. The project aims to provide a framework for training generic object extraction operators from small sets of training data based on extensions to fully Bayesian approaches to object categorisation. The techniques developed are designed to take the statistical properties of SAR, polarimetric SAR and interferometric SAR into account. To validate the proposed approaches, operators for the extraction of individual buildings and parts of road-networks in urban areas will be trained. The data required for this evaluation consists of high resolution imagery, single and dual pol, of an urban area. The project is funded by the German Research Foundation (DFG). The proposal concerned was recently approved, and a PhD position will be available for supporting research in this project.

Back to list of proposals

DLR 2004-2016