GeoSAR P-band and X-band Performance In Southern California and Colombia, South America

Transcription

1 GeoSAR P-band and X-band Performance In Southern California and Colombia, South America ISPRS International WG 1/2 Workshop 2005 James J Reis, EarthData Technologies Dr. Scott Hensley, Jet Propulsion Laboratory 6-10 June 2005 ISPRS Conference, Bamff, Canada 1

2 Outline GeoSAR Overview, Including Latest Upgrades NOAA Watershed Project Overview Planarmetric Accuracy (X-band) Effects of Notching P-band Transmit Waveform Anecdotal P-band Ground Penetration of Arid Soil GeoSAR Results in Colombia South America X-band vs. P-band Reflectance Differences Foliage Penetration (Early Results) GeoSAR Collection and Processing Methodology Inherent Collection Redundancy Mosaic of Multiple Looks Issues Associated with Differential Penetration 6-10 June 2005 ISPRS Conference, Bamff, Canada 2

3 GeoSAR Overview, Including Latest Upgrades 6-10 June 2005 ISPRS Conference, Bamff, Canada 3

4 Aircraft Exterior Two P-band Antennas Are Mounted On Each Wingtip And Operate Over Mhz 20m40m Dual Baseline Two X-band Antenna Are Mounted Under Each Wing Close To The Fuselage And Operate Over Mhz New 2 nd Generation Antenna Position Measurement Unit 6-10 June 2005 ISPRS Conference, Bamff, Canada 4

5 GeoSAR Data Collection X-Band (3 cm wavelength) Collection Height: 33,000 to 39,000 ft MSL P-Band (85 cm wavelength) (New) Profiling LIDAR (3 Returns, 3 Intensities) 12km 10.5 km 12km Simultaneous X-band and P-band Interferometric Radar Data Collection on Left and Right Sides of Aircraft Flight Lines up to 1200 Km in Length, 380MBs Data Rate Flight Duration Approximately 5 Hours 6-10 June 2005 ISPRS Conference, Bamff, Canada 5

6 GeoSAR Redundant Collection Methodology Collection Redundancy Project Boundary Opposite Side Looks on Project Boundary Flight Direction (Looking Down) Port Look Nadir Hole Starboard Look GeoSAR Swath Coverage Redundant Looks 6-10 June 2005 ISPRS Conference, Bamff, Canada 6

7 User User Request Request The Basic GeoSAR Process Flow ACQUISITION PROCESS PRODUCTION PROCESS FINISHING PROCESS FPS Flight Planning System Data Collection Ground Processor Mosaic & Merge DEM Editing Radar Command Disk Flight Log Radar Data dgps Data Formatting & Distribution Navigation Software GulfStream II NavCom Real Time GPS Mission & Ground Crew Air Traffic Control Frequency Permits Gnd Station Differential GPS Ground Reference (Optional) Customer Data Lidar Profiler Data 6-10 June 2005 ISPRS Conference, Bamff, Canada 7

8 LIDAR Profiler 3 Returns, 3 Intensities LIDAR Profiler Add a lidar profiler to operate at high altitude Reduces/eliminates need for ground control in remote, inaccessible or inhospitable areas. Improves accuracy of mosaicked products and DEM. Calibrated Profiles Overlaid on Reference DEM 98% 1 st Returns 27% 2 nd Returns Leica ALS40 Lidar Modified for High Altitude Profiling 6-10 June 2005 ISPRS Conference, Bamff, Canada 8

9 Profile of Hansen Dam 2-3m Spot size 3-5cm Spot Posting 3 Returns 3 Intensities 6-10 June 2005 ISPRS Conference, Bamff, Canada 9

10 Application Of Lidar Profile To Corrective Image/DEM/Mosaic Formation Swath Is Flow With SAR Lidar Cross Swath Is Flown Lidar Profile Correlated With The Swath Used in Mosaic Level Affine Adjustments 6-10 June 2005 ISPRS Conference, Bamff, Canada 10

11 Synopsis of GeoSAR Features Dual-band X & P, Dual Sided, Single Pass Interferometric 4Q05 Upgrade: 10-bit Data Samples (no compression) Dual Baseline X-band 370MBS collection Rate Two Sided Systems (GeoSAR) Collects Twice As Much Data Per Flight Distance Flown The Only Commercial IFSAR With A Co-mounted Lidar Profiler GeoSAR Flightline Spacing Yields An Additional Factor Of 2 Increase In Collection Data Redundancy LMS Mosaic Error Propagation Improves DEM Quality Without Effecting Spatial Resolution (MTF) 6-10 June 2005 ISPRS Conference, Bamff, Canada 11

12 NOAA Watershed Project Overview Emphasis on P-band 6-10 June 2005 ISPRS Conference, Bamff, Canada 12

13 NOAA Project, All Phases 315 Quads Delivered, 3m Post, 1m Vertical RMSE 6-10 June 2005 ISPRS Conference, Bamff, Canada 13

14 Problematic Measurement of Planarmetric Accuracy from Surveyed Point Location of a Survey Point to SubPixel Accuracy is Not Possible Description: NW Corner Of Parking Area On North Side Of Morales Street, Just East Of Hubbard Avenue 6-10 June 2005 ISPRS Conference, Bamff, Canada 14

15 Extended Control Feature Extraction Process ADS40 1ft TruOrtho Imagery 1 GeoSAR 3m X-band Imagery Identify Extended Feature Extract Extended Feature Vector Register Feature to X- band Imagery Label Unique Pattern Handle 6 Pattern Handle Location Information 6-10 June 2005 ISPRS Conference, Bamff, Canada 15

16 Ortho Control (Pattern-2) The closed loop patterns have high radar contrast that create many simultaneous spatial constraints. Vector Pattern Best Match ADS40 1ft GSD Natural Color TruOrtho 3m GSD X-band Imagery Blue Lines = Ortho Reference (Control) Red Lines = Best Translational Match (Radar) 6-10 June 2005 ISPRS Conference, Bamff, Canada 16

17 ADS40 1ft GSD Natural Color TruOrtho 3m GSD X-Band imagery Vector Pattern Best Match Blue Lines = Ortho Reference (Control) Red Lines = Best Translational Match (Radar) Ortho Control (Pattern-7) 6-10 June 2005 ISPRS Conference, Bamff, Canada 17

18 Scatter Plot of Residual Errors (Bias Removed) Scatter Plot of Residuals Mean(x,y) = (2.35,-1.05) Y-Displacement (m) X-Displacement (m) 6-10 June 2005 ISPRS Conference, Bamff, Canada 18

19 P-Band Frequency Management 315 Quads of Collection Area Preprogrammed waveform is notched to assure compatibility with other users of MHz frequency band. The greater the notching in an area, the lower the resolution of the P-band image; some degradation of P- band DEM. Optimal location for P- band processing is in an area where frequency usage is lowest June 2005 ISPRS Conference, Bamff, Canada 19

20 NOAA-1 P-band Transmit Waveform (Frequency Domain) Non-Notched Spectral Shape 0 NOAA-1 P-band Waveform (Frequency) Many Discrete Frequency Notches Justice Dept. Notch db MHz Bathtub Notch June 2005 ISPRS Conference, Frequency Bamff, (MHz) Canada 20

21 GeoSAR P-band Image, 5m Posting (Zoomed) Note: Imagery Is Blurred In The Range Direction Due To High Sidelobe Structure In The Transmit Waveform Increasing Range 6-10 June 2005 ISPRS Conference, Bamff, Canada 21

22 X-band Drainage (3m Posting) X_BAND_DRAIN.TIF 6-10 June 2005 ISPRS Conference, Bamff, Canada 22

23 P-band Drainage (5m Posting) P_BAND_DRAIN.TIF 6-10 June 2005 ISPRS Conference, Bamff, Canada 23

24 Linear Feature Orthophoto 6-10 June 2005 ISPRS Conference, Bamff, Canada 24

25 Linear Feature X-band vs. P-band Linear Features X_BAND_LINEAR_FEATURE_MAG.TIF P_BAND_LINEAR_FEATURE_MAG.TIF 6-10 June 2005 ISPRS Conference, Bamff, Canada 25

26 Band 742 Landsat Merged with 4m Aerial Photography Complements of Dr. Martin Insley, Senior Geologist, Infoterra Ltd. (Used with Permission) 6-10 June 2005 ISPRS Conference, Bamff, Canada 26

27 Geological Formation X-band 3m Posting Edwards_X_dem2.tif Edwards_X_mag2.tif Direction of Illumination 6-10 June 2005 ISPRS Conference, Bamff, Canada 27

28 Geological Formation P-band 5m Posting Edwards_P_dem2.tif Edwards_P_mag2.tif Direction of Illumination 6-10 June 2005 ISPRS Conference, Bamff, Canada 28

29 Profile of Geological Formation Height (m) Direction of Illumination Profile Comparison (X vs. P) NOAA_P_profile (5m posts) NOAA_X_Profile (3m Posts) Distance (m) 6-10 June 2005 ISPRS Conference, Bamff, Canada 29

30 Comparison Between P, X and X-P Band Surfaces Negative values indicate areas where P band is higher than the X band elevation Positive values indicate areas where P band is lower than the X band elevation I.e. areas with ground penetration X P Difference Surface X Band Surface P Band Surface Complements of Dr. Martin Insley, Senior Geologist, Infoterra Ltd. (Used with Permission) 6-10 June 2005 ISPRS Conference, Bamff, Canada 30

31 Comparison Between X, P and X - P Vertical Profiles Complements of Dr. Martin Insley, Senior Geologist, Infoterra Ltd. (Used with Permission) 6-10 June 2005 ISPRS Conference, Bamff, Canada 31

32 GeoSAR Results in Colombia South America 6-10 June 2005 ISPRS Conference, Bamff, Canada 32

33 Alpha Collection Pipeline Terminus X-band, 3m Post 6-10 June 2005 ISPRS Conference, Bamff, Canada 33

34 Alpha Collection Pipeline Terminus P-band, 5m Post Wave Action over 90 Second Aperture Time Distorts Pier Mulipath 6-10 June 2005 ISPRS Conference, Bamff, Canada 34

35 X-band P-band 3M 5M DEM 6-10 June 2005 ISPRS Conference, Bamff, Canada 35

36 X-band vs. P-band DEM Comparison X-band profile 6-10 June 2005 ISPRS Conference, Bamff, Canada 36

37 Pipeline Terminus, Colombia P-band, X-band, 3m 5m post 6-10 June 2005 ISPRS Conference, Bamff, Canada 37

38 P-Band Magnitude Image 6-10 June 2005 ISPRS Conference, Bamff, Canada 38

39 X-band 3m Posts XR_Line61_zoom.tif 6-10 June 2005 ISPRS Conference, Bamff, Canada 39

40 P-band 5m Posts UR_Line61_zoom.tif 6-10 June 2005 ISPRS Conference, Bamff, Canada 40

41 GeoSAR Collection and Processing Methodology 6-10 June 2005 ISPRS Conference, Bamff, Canada 41

42 GeoSAR Redundant Collection Methodology Collection Redundancy Project Boundary Opposite Side Looks on Project Boundary Flight Direction (Looking Down) Port Look Nadir Hole Starboard Look GeoSAR Swath Coverage Redundant Looks 6-10 June 2005 ISPRS Conference, Bamff, Canada 42

43 Example of Textural Banding in Nexus Data Unfiltered Spatial Filter Filtered 6-10 June 2005 ISPRS Conference, Bamff, Canada 43

44 P Band Slope Map Single-Look DEM Two-Look DEM Complements of Dr. Martin Insley, Senior Geologist, Infoterra Ltd. (Used with Permission) 6-10 June 2005 ISPRS Conference, Bamff, Canada 44

45 Radar Intensity Data Amplitudes (I,Q) Complex RF Signal Amplitude Has ~ Normal Distribution One RF Cycle (Degrees) I Q Probabality (X) Normal Probability Density Function (Amplitude of RF varies Gaussian) Intensity Value (X) Real Channel I Σ Quadrature Channel ( Q ) Magnitude I + Q 2 2 Rayleigh Probability Density Function of The Magnitude Signal 6-10 June 2005 ISPRS Conference, Bamff, Canada 45 Probabality (X) Intensity Value (X) Rayleigh

46 Comparison of Intensity Distributions Probability Density Functions (PDFs) Probabality (X) Rayleigh Normal Intensity Value (X) Optical Imagery has 2 Degrees of Freedom: Gain (Standard Deviation, σ) Brightness (Mean, µ) Normal Distribution (Optical): 2 2 { ( ) } 1 F( x) = exp x µ / 2 σ, σ 2π µ =3.0 and σ=0.66 Radar Imagery has 1 Degree of Freedom: Scaling Parameter, β Rayleigh Distribution (Radar): x 2 2 F( x) = exp 2 ( x / 2 β ), β=1.0 β 6-10 June 2005 ISPRS Conference, Bamff, Canada 46

47 Affects of Additional Looks 1st Look Mean =0.13 Stdev = 0.07 Forecast: Single Look 100,000 Trials Frequency Chart 98,438 Displayed nd Look ½ Σ Mean =0.13 Stdev = 0.05 Forecast: Double Look 100,000 Trials Frequency Chart 98,925 Displayed Mean =0.13 Stdev = /3 Σ Forecast: Triple Look 100,000 Trials Frequency Chart 99,105 Displayed rd Look June 2005 ISPRS Conference, Bamff, Canada 47

48 Histogram Comparisons 100,000 Trials Frequency Chart 98,438 Displayed Look Forecast: Single Look 2 Looks Looks As The Number Of Looks Increase, The Histogram Takes On A Decidedly More Gaussian Shape The Greatest Change Occurs from 1 to 2 Looks 3 Or More Looks Have Minimal Affect On Output Texture 6-10 June 2005 ISPRS Conference, Bamff, Canada 48

49 Signal Processing Issues In general for radar imagery, removal of textural banding in the mosaic process by gain adjustments also leads to a contrast differential (and vice versa), which in turn causes brightness banding One solution is to filter the data spatially prior to mosaicking so the amplitude statistics for adjacent regions are (more or less) identical. This, however, will show up as a loss of spatial detail in the filtered areas EarthData does not spatially filter its radar products, preferring instead to deliver to the client products with full spatial resolution over products with cosmetic improvements in viewing quality 6-10 June 2005 ISPRS Conference, Bamff, Canada 49

50 Questions? 6-10 June 2005 ISPRS Conference, Bamff, Canada 50