Data Processing Developments at DFD/DLR Stefanie Holzwarth Martin Bachmann, Rudolf Richter, Martin Habermeyer, Derek Rogge EUFAR Joint Expert Working Group Meeting Edinburgh, April 14th 2011
Conclusions (Session 12) The systematic collection of data asks for standards in preprocessing of the data, selection and filtering of useful information from whole acquisition. Putting things together into a commonly usable (and effectively USED) storage system should be followed up by the big players... kindly provided by D. Schläpfer
OpAiRS Processing Chain L0 - Raw Data Data Transcription Build Metadata Data QC QualityControl L0+ Laboratory Calibration Level 0+ Product Archiving L0+ Vicarious Calibration Onboard Calibration Sources System Correction Data QC Level 1 Product QualityControl L1 Archiving L1 Attitude Data, Position Data, DEM Parametric Geocoding Data QC Radiative Transfer Model, Meteorologic Data, DEM Level 2a Product Atmospheric Correction Data QC QualityControl L2 Level 2 Product Folie 3 Archiving L2
EnMAP Processing Chain Raw data Orbit and Attitude Products Level 0 Processor Transcription Long Term Archive L0 Data In-flight Calibration Calibration Products Data QC Implementation: Automated within processors QC-related Metadata Level 1 Processor Systematic and Radiometric Correction Data QC routines Level 2geo Processor Orthorectification O U T P U T L1 Product L2geo Product Interactive procedures Data QC routines Level 2atm Processor Atmospheric Correction Data QC routines P R O C E S S O R L2 Product Interactive Data QC for selected scenes on a regular basis Data QC reports L2atm Product Folie 4
Automated Data Quality Control Including Quality Indicators (QI) for General sensor characterization (e.g., spectral smile) Sensor calibration issues (e.g., striping in pushbroom sensors) Sensor performance during data acquisition (e.g., data drops) External conditions during overflight (e.g., cloud coverage) Processing (e.g., uncertainty of geo-location) Quality of auxiliary data used in processing (e.g., DEM accuracy) Folie 5
HDF5-Design of Level2 Data Folie 6
ATCOR4 updates Recompiled with MODTRAN 5 code using 0.4nm (instead of 0.6nm) Fontanala-2011 solar irradiance spectrum included Improved spectral smile tool Removal of haze over water Derivative filter for spectral polishing Sensor definition panel Folie 7
Automated calibration site locator Spectral and spatial approach: Spectral: surface materials that have minimal spectral features. Spatial: region that are spatially homogeneous. Temporal (if available): surface material that do not change properties from day to day or season to season. Methodology: Spectral: applies a spatial boxcar smoothing filter (e.g. 7 bands) and calculates the difference from the original image. Spatial: applies a sliding window spatial mean filter (e.g. 7x7 window) and calculates the difference from the original image. Temporal: determine regions that show the minimal spectral changes from one period to another. Combine outputs from spectral, spatial and temporal data. Folie 8
Oberpfaffenhofen calibration test site True Colour RGB Spectral Difference Spatial Difference Combined Difference Folie 9
Oberpfaffenhofen calibration test site True Colour RGB Spectral Difference Spatial Difference Combined Difference Folie 10
Oberpfaffenhofen Temporal Data Integration 030630 040520 050603 090722 Combined temporal results Folie 11
Oberpfaffenhofen Temporal Data Integration 030630 040520 050603 090722 Combined temporal results Folie 12
Spatial-spectral bad-pixel masking Removal of bad pixels (+noise) via integration of: Spectral (band to band correlation) Drawback: deep narrow absorption bands Spatial (pixel to pixel correlation) Drawback: edge detection and point targets Combined approach reduces false positives Folie 13
Spatial-spectral bad-pixel masking Spatial-spectral removal of bad pixels (noise) Spectral metrics: RMS Spectral Angle Integrated results Spatial metrics: RMS Spectral Angle Folie 14
Airborne Data and Line Leveling AISA, 400-2500nm, 178 bands, ATCOR4 Atmospheric Correction Data owned by Goldbrook Ventures, Vancouver B.C. Flight lines average overlap 30% 2 meter pixels, flown in late July 2008 Code developed at the University of Alberta, Canada Rivard, B.; Rogge, D.; Feng, J. Folie 15
Leveling Results Code developed at the University of Alberta, Canada Rivard, B.; Rogge, D.; Feng, J. Folie 16
Conclusions Monitoring of sensor performance and data quality is an essential part of our processing chain Harmonization and standardization in progress EUFAR guidelines for Quality Layers and extended metadata CEOS QA4EO guidelines HDF format for compilation of data Operational QC to be extended and harmonized Development of data QC for thematic products Spatial-spectral bad-pixel masking Automatic calibration site locator for in-flight calibration Line leveling: Adapt code to DLR s processing chain Folie 17