Land Cover and Asset Mapping Operational Change Detection Andreas Müller DLR DFD with the support of Charly Kaufmann, Allan Nielsen, Juliane Huth ESA Oil & Gas Workshop, 13-14 September 2010 Folie 1 TWOPAC - Automated Land Cover Classification
Prerequisits for Monitoring Systems Demand Reliable Data and guaranteed access Optimized data handling Operational robust algorithms Standardised procedures Fast information access tailored to user needs Slide 2
The Demand In 2007, the European Commission Vice President, responsible for Industry policy declared that European industries need predictability in the flow of raw materials and stable prices to remain competitive. We are committed to improve the conditions of access to raw materials, be within Europe or by creating a level playing field in accessing such material from abroad. Slide 3 3
International framework G8 Summit June 2007: Declaration on "Responsibility for raw materials: transparency and sustainable growth", which addresses the key priorities for a sustainable and transparent approach. Federal Government 2007: Raw materials strategy in compliance with different aspects of EITI and to support raw materials investment abroad. European Commission 2008: The Raw Materials Initiative Meeting our Critical Needs for Growth and Jobs in Europe (COM(2008)699). EITI Extractive Industries Transparency Initiative: Setting global standards for transparency in oil, gas and mining. GEO Group on Earth Observations: The responsible management of Earth s environment is one of today s most pressing concerns and a central motivation for the Group on Earth Observations. Slide 4
Linking politics to Earth observation EO-MINERS aims at integrating new and existing Earth Observation tools to improve best practice in mining activities and to reduce the mining related environmental and societal footprint by introducing innovative remote sensing tools to the mining industry, providing accuracy and quality measures for remote sensing products, demonstrating the application of Earth Observation in different case studies, fostering the dialogue between mining industry and environmental organisations based on EO-derived information and generalising the obtained results to be used in operational mining applications in the future. Slide 5
EO Application and Development using Demonstration Sites Policy Analysis, Requirements Definition EO-based Information System Standards & Protocols Application Development Integration & Products Communication, Dissemination, Capacity Building Slide 6
Standards & Protocols DLR is leading a WG supported by the Assoc. of German Engineers (VDI) within the network of excellence of the BMBF for Optical Technologies: Earth Observation by Imaging Spectroscopy using Airborne Sensor Systems guideline is a pre-stage of a DIN / ISO EN, but accepted by legislative overlap with different standardization activities in Photogrammetry and RS in Germany => DIN ISO 18717, 18740, etc., including standards of the ISO TC 211 regulations Similar activity in Europe on airborne hyperspectral sensor systems Slide 7
Entity Airborne Remote Sensing (E.ARS) ISO9001:2000 certified http://www.opairs.aero http://cocoon.caf.dlr.de http://atcor.dlr.de http://www.dlr.de/fb Campaign information and survey requests, information on sensors, laboratories and processing DLR Spectral Archive - incl. processing tools ATCOR - Atmospheric Correction Software DLR flight facilities Slide 8
Geometric Corrections (ORTHO) (1/2) Scene (original) Reference image database Digital elevation model database Extraction of ground control points by matching Improvements of Line-of-Sight Vector Generation of ortho image Scene (after geometric correction) Slide 9
Geometric Corrections (ORTHO) (2/2) European Mosaic (GMES Image 2006, ESA-Project) Proj.: European, Res.: 25m, Acc.: 10m; based on ~1500 SPOT + IRS scenes Slide 10
ATCOR : Atmospheric & Topographic Correction Example of Cloud Shadow Removal HyMap, Chinchon, Spain, 12 July 2003 RGB=878/646/462 nm Slide 11
Atmospheric Correction: Haze Removal Location: near Rostock, Germany True color image from: ALOS/AVNIR-2 Scene (original) Scene (after atmospheric correction) (including haze correction) Slide 12
Data and classification scheme Multispectral, mono-temporal data of different spatial resolutions Based on FAO LCCS scheme comparability of results Hierarchical structure e.g. Separation of 5 levels 3 FAO-LCCS Levels, 2 Detail Levels in coordination with Vietnamese partners Classification scheme can be adjusted or replaced 2.4m 10m 30m Slide 13
Automated Process Chain TWOPAC for classification of multi-spectral data TWOPAC Twinned object and pixelbased automated Classification chain Processing chain Modular Processing: Pre-Processing Steps Database for sample data management Different classification methods implemented (C5.0, SVM, Maximum likelihood) Validation and iterative classification Independent from sensor type Processing of mono-temporal, multispectral data Slide 14
Classification process: Classification procedure II INPUT DATA PRE- PROCESSING Classification of input image either pixel-based or object-based CLASSIFICATION SCHEME SAMPLING Generation of Metadata for integration into ELVIS Information System SAMPLING DATABASE EXPORT TRAINING VALIDATION Classified Image CLASSIFIER CREATION Confusion Matrix Classification result Integration in ELVIS Information System Slide 15
Remote sensing products for ELVIS WISDOM Information System as an instance of ELVIS Land cover information as base dataset for several applications Slide 16
Automated Multivariate change detection Multi- or hypervariate data covering the same geographical region recorded at n- different points in time Supervised/unsupervised methods Two major families for unsupervised Segmentation/clustering/classification at each point in time, look for changes in classification results Detect change directly in the measured variables or derived features, pixelbased (scale-space, object-oriented ) Future Perspective: unsupervised, direct methods only, fully automatic Allan A. Nielsen, K. Conradsen and James J. Simpson (1998). Multivariate Alteration Detection (MAD) and MAF Post-Processing in Multispectral, Bi-temporal Image Data: New Approaches to Change Detection Studies. Remote Sensing of Environment 64, 1-19. Slide 17
MAD, data example Landsat-5 TM data from Hanford, Washington, USA 12 Aug 1983 and 28 Jun 1987 850 x 750 pixels (TM1 and 6 left out) Coregistration to rms-error of around 0.5 pixel Changes due to excavation and exploratory activities for waste disposal Slide 18
Hanford, TM TM bands 7, 4 and 2 28 Jun 1987 TM bands 7, 4 and 2 12 Aug 1983 Slide 19
Hanford, Change Detection Sum of standardized, squared MADs Slide 20
Reliable Data Access Sentinel 2 Multi-spectral optical imager 13 spectral channels VIS, NIR, SWIR 2 satellites High repetition rate Easy open access Operational Services Philippe Martimort, Michael Berger, Bernardo Carnicero, Umberto Del Bello, Valérie Fernandez, Ferran Gascon, Pierluigi Silvestrin, François Spoto & Omar Sy Olivier Arino, Roberto Biasutti & Bruno Greco,, Sentinel 2, the optical high resolution component for GMES operational Services ESA bulletin 131, 2007 Slide 21