FD-3: Semantic content extraction from high resolution EO images
Sunday, July 6, 08:30 - 17:30
Mihai Datcu, German Aerospace Center, DLR Oberpfaffenhofen and Klaus Seidel, Swiss Federal Institute of Technology, ETH Zurich
Adding meaning to EO images is an important practical problem raising many theoretical challenges. The functional applications are image annotation, indexing, understanding, data mining, or integration in GIS.
EO images with resolutions in the range of 0.5m to 2m, of natural and mainly of sites or objects related to human activities, are difficult to interpret since they make 3D structures evident, thus increasing the image complexity. Their understanding is strongly depending on the spatial contextual information, and also conditioned by the interpretation context, thus a domain ontology is often needed.
Meanwhile, users have at their disposition tools for the definition of specific goals using semantics. The extraction of image meaning is based on the principle of semantic compositionality: the meaning of a whole is a function of the meanings of its parts and their mode of syntactic combination. Thus, an image has to be decomposed into atomic cues or signs, and semantics will be induced in an interactive learning process grouping existing image structures in configurations, considering their categories, and generalizing results over an entire image.
These are difficult tasks requiring cooperative solutions integrating a variety of methods of soft computing, information semantics and the semantic web, advanced statistics, saliency and latency analysis, and probabilistic reasoning. The final goal is to design machines more closely interacting at human conceptual levels thus to help automation of the human EO image analyst work, as Image Information Mining systems, which can be operated using intelligent interfaces able to correlate the information content of the images with the relevant goals of the application.
- Basics of image formation for SAR and optical sensors
- SAR and optical image content modeling, and extraction of cues and primitive features
- Image salient feature extraction: statistical, gestalt and information theory
- Grouping, and object categories definitions: clustering, graph analysis and image topology
- Latent Semantic Analysis: the basic statistical approaches
- Machine Learning: on the use of neural networks, SVM, and Bayesian methods
- Image semantics and knowledge representation
- Image Information Mining systems
All chapters will be illustrated by examples and demos using high resolution SAR and optical images in relevant case studies: risk and hazard, pattern recognition, cartography, semantic indexing, etc.
Mihai Datcu received the M.S. and Ph.D. degrees in Electronics and Telecommunications from the University “Politechnica” of Bucharest UPB, Romania, in 1978 and 1986. In 1999 he received the title “Habilitation à diriger des recherches” from Université Louis Pasteur, Strasbourg, France. He holds a professorship in electronics and telecommunications with UPB since 1981. Since 1993 he is scientist with the German Aerospace Center (DLR), Oberpfaffenhofen. He is developing algorithms for model based information retrieval from high complexity signals and methods for scene understanding from synthetic aperture radar (SAR) and interferometric SAR data. He is engaged in research related to information theoretical aspects and semantic representations in advanced communication systems. Currently he is Senior Scientist and Image Analysis research group leader with the Remote Sensing Technology Institute (IMF) of DLR, Oberpfaffenhofen, coordinator of the CNES-DLR-ENST Competence Centre on Information Extraction and Image Understanding for Earth Observation, and professor at ENST Paris. His interests are in Bayesian inference, information and complexity theory, stochastic processes, model-based scene understanding, image information mining, for applications in information retrieval and understanding of high resolution SAR and optical observations.
He has held visiting professor appointments from 1991 to 1992 with the Department of Mathematics of the University of Oviedo, Spain, from 2000 to 2002 with the Université Louis Pasteur, and the International Space University, both in Strasbourg, France. In 1994 was guest scientist with the Swiss Center for Scientific Computing (CSCS), Manno, Switzerland and in 2003 he was visiting professor with the University of Siegen, Germany. From 1992 to 2002 he had a longer invited professor assignment with the Swiss Federal Institute of Technology ETH Zürich. He is involved in advanced research programs for information extraction, data mining and knowledge discovery and data understanding with the European Space Agency (ESA), Centre National d’Etudes Spatiales (CNES), NASA, and in a variety of European projects. He is member of the European Image Information Mining Coordination Group (IIMCG).
Klaus Seidel received his B.S. degree in experimental physics in 1965 and the Ph.D. degree in 1971, both from the Swiss Federal Institute of Technology (ETHZ). He was with the Computer Vision Lab at ETHZ and head of the remote sensing group until 2002. Since 1987 he was a Swiss Delegate and Expert in various ESA Working Groups and at the same time functioning as the National Point of Contact for Switzerland. He is currently consultant for ESA projects specialized in image information mining related to remote sensing archives. He was also teaching courses in digital processing of satellite images and has published several papers concerning snow cover monitoring, geometric correction and multispectral analysis of satellite images, and on remote sensing image archival. Most recently he published a book together with J. Martinec on “Remote Sensing in Snow Hydrology”.
He was involved in the Knowledge-driven Image Information Mining (KIM) project and is currently involved in the Knowledge Enabled Services KES, KIM Validation and Knowledge Centered EO (KEO) projects for ESA. Dr. Seidel is member of the European Image Information Mining Coordination Group (IIMCG) and the Data Archiving and Distribution Technical Committee of IEEE Geoscience and Remote Sensing Society.