DEM Processing Chain & Data Products

Transcription

1 DEM Processing Chain & Data Products Birgit Wessel & Mosaicking and DEM Calibration Team German Remote Sensing Data Center (DFD-DLR) TanDEM-X Science Team Meeting 2008-Nov-24

2 DEM Processing Chain InSAR: Integrated TanDEM-X Processor (ITP) DEM Mosaicking And Calibration Processor (MCP) DEM Generation TanDEM-X DEM Raw DEM: elevation Height Error Map Amplitude Flag Mask Preparation Calibration Detection of water bodies Mosaicking Quality Control mosaicked DEM: elevation Height Error Map Amplitude Mosaic Flag Mask Slide 2

3 Global coverage strategy for deriving a DEM Long continous DEM acquisitions Two coverages of the earth surface Red: 1st year coverage green: 2nd year Slide 3

4 MCP Production Phases Data-driven Preparation Acquisitions Phase 1 ~ 0,4 TBytes/day = 400 scenes / day 1st year DIMS: Archive Raw DEM P1 Preparation Processor water body detection Tie-point candidates Height discrepancies pot. WAM 1 TP chips DB residues reference Tie-pointing: Scene based DEM Calibration Mosaicking: Scene based Slide 4

5 Detection of Water bodies Basic approach: Threshold method Search restriction: exclusion of polar and desert regions due to MODIS classification Extended approach: Use of existing water masks like SRTM, GSHHS and refine the outlines (active contours) Use of additional features (texture) Flag mask contains water bodies, shadow and layover areas, and probability levels for water DEM values of water bodies won t be modified Limitations: automatic approach Lake Lucerne (Switzerland), TSX-VV Slide 5

6 DEM Production Phases - Intermediate DEM Acquisitions Phase 1 DIMS: Archive Raw DEM P1 1st year corrections Raw DEMs P1 2nd year ~ 0,4 TBytes/day = 400 scenes / day Intermediate DEMs Preparation Processor Calibration Processor water body detection Tie-point candidates Height discrepancies Block adjustment DEM verification DEM fusion Intermediate DEM of 1st year acquisitions Mosaicking Processor Tie-pointing: Scene based pot. WAM 1 TP chips DB residues reference DEM Calibration Mosaicking: Scene based Slide 6

7 DEM Calibration Height error model DEM Calibration: Correction of each DEM acquisition due to systematic residual errors g HEIGHT ERROR MODEL 2 3 n ( x, y) = an + bn x + cn y + dnxy + en y + fn y Main contribution: a-c n index of the data take a n...f n unknown parameter x,y range and azimuth a: absolute height offset (~1.5m) b: slope in range (~4mm/km) c: slope azimuth (~0.3mm/km) Corrected Height: H = H_InSAR + g(x,y) DEM Azimuth (y) Range (x) Slide 7

8 DEM Calibration Reference Data (1) Calibration: Estimation of correction parameters by Least-squares adjustment Tie-Points Ground Control Points Tie-Points Identical points in overlaps Height differences to other acquisitions are used for the adjustment Filtering to reduce noise (2m-> e.g. 0.7m) No absolute height information Slide 8

9 DEM Calibration Reference Data (1) Calibration: Estimation of correction parameters by Least-squares adjustment Tie-Points Ground Control Points Ground Control Points ICESAT Coverage Use of ICESat data Estimation of height offsets to WGS84 Height differences to reference data GCP source Coverage Accuracy ICESat Global 0.1m 1(2)m (weather/terrain) Slide 9

10 Example: DEM Calibration of E-SAR DEM data E-SAR team E-SAR team Slide 10

11 Difference between E-SAR DEM and Laser DEM LVG Bayern E-SAR team Slide 11

12 Difference between calibrated E-SAR DEM and Laser DEM LVG Bayern LVG Bayern Differences in [m]: before and after DEM Calibration Slide 12

13 Difference between E-SAR DEM and GPS 1. und 2. order deformations are corrected After calibration: In the northern part: relative height error < 2 m height differences < 10 m, higher differences are based on GPS difficulties Differences in [m]: before and after DEM Calibration Slide 13

14 DEM Verification & Validation Concept Verification after DEM Calibration (during processing) GPS-Tracks: Use of kinematic GPS transects in east<- >west direction for long error characteristic s and for relative error verification Build up of a science team for the evaluation activities Planned GPS-Tracks World-wide Slide 14

15 Mosaicking / DEM Fusion Applies estimated systematic corrections to the Raw DEMs Mosaic generation: Averages all available Raw DEMs to minimize random error Followed by operator-conducted QC DEM Slide 15

16 Block 0256 Mosaicking Quality Control Region Balkan Num. of RawDEMs 54 Parameter Pool Log File Quality APPROVED Proc. Date/Time :13: Mosaic Region status Quicklook set status flag set emergency reprocessing flag for selected RawDEMs DEM HEM AMP OPER_0256_54 Balkan APPROVED OPER_0256 DIFF COV WAM OPER_0257_26 Xinjiang APPROVED Difference +-10 m Ref DEM - InSAR OPER_0258_39 Detailed Greece NOT_APPROVED differences OPER_0259_28 Xinjiang APPROVED Statistic: valid pixel pairs checked abs Maximum value of difference Minimum value of difference Mean value of difference 1.14 Std deviation of difference 4.20 RMS of difference 4.35 Slide percent of data within +-2.0, 90 percent within +-5.0

17 DEM product: TanDEM-X DEM The DEM product consists of the following information layers: Elevation (DEM, 4 Byte) Height Error Map (HEM, 4 Byte) Flag Mask (FLM, 1 Byte) Water body flags Shadow and layover flags Optional: Amplitude Mosaic (AMP, 2 Byte) DEM HEM TanDEM-X DEM -> 4 years after launch (-> Oct. 2013) Intermediate TanDEM-X -> 2,5 years after launch with first global coverage (less quality) Tiles in Geotiff in 1 x 1 in geogr. coordinates FLM AMP Slide 17

18 DEM Products: DEMs on special user request FDEMs: DEMs processed to finer pixel spacing and higher random height error HDEMs: HRTI-4 like DEMs. Additional DEM acquisitions needed. Strongly limitated! Processing after all standard TanDEM acquisitions are made and calibrated. Local DEMs. Slide 18

19 TSX TDX TanDEM-X TerraSAR add-on for Digital Elevation Measurements Thank you for your Attention! Slide 19