dtm extraction beneath canopy using l- and p...

Post on 29-Jun-2020

2 Views

Category:

Documents

0 Downloads

Preview:

Click to see full reader

TRANSCRIPT

© 2006 Intermap Technologies, Inc. All rights reserved.

DTM Extraction Beneath Canopy Using L- and P-Band Data from INDREX-II:

Preliminary Results

Presented to the PolInSAR Workshop: Frascati, ItalyJanuary 24, 2007

Bryan Mercer, Qiaoping Zhang

Intermap Technologies Corp.

Calgary, Canada

Acknowledgements:

ESA (Malcolm Davidson), DLR, WUR / SARVision

Alberta Ingenuity Fund

© 2006 Intermap Technologies, Inc. All rights reserved.

Contents

IntroductionRationale and Objectives

The Approach

Current results

Issues to be Resolved

© 2006 Intermap Technologies, Inc. All rights reserved.

Rationale and Objectives

Extraction of ground elevation (DTM) beneath canopyUseful accuracy desired ~ 2-3 meters RMSEPrevious P-Band experience with TopoSAR systemHow does L-Band performance compare?

Opportunity, through ESA, to use INDREX-II data setL- and P-Band quad-pol, interferometric dataGround truth

Objectives of this activity:To compare DTM performance of P- and L-Band To understand the issues and sensitivities with respect to various factors such as forest height and density, slope etcCurrently at an early stage

© 2006 Intermap Technologies, Inc. All rights reserved.

Previous Work with TopoSAR SystemCapitol Forest Elevation Profiles

145

150

155

160

165

170

175

180

185

0 100 200 300 400 500 600 700 800

Cross SectionM

eter

s

Meters

LIDAR ‘Truth’ P-Band

X-Band

30 m

© 2006 Intermap Technologies, Inc. All rights reserved.

The Approachassume RVOG* model applicable,

extract ground phase Φ

* Treuhaft and Siqueira (2000), Papathanassiou and Cloude (2001)

Real

Imaginary

+*

ΦxΦ = κZ Z0

= Ground Phase

= ground/volume ratiois polarization dependent

© 2006 Intermap Technologies, Inc. All rights reserved.

The Reality

γobs = γv γsnr γt γrgsp ...~ ~ ~

INDREX-II Final Report to ESAIrena Hjansek, DLRDirk Hoekman, WUR

≈ 1 We hope!

© 2006 Intermap Technologies, Inc. All rights reserved.

Process Summary

L- and P- BandCo-registered SLC pairs

(plus ancillary info)

Coherence Optimization

Coherence

X-BandGeocoded

DSM and image

GPS GroundElevations

CorrectionSurface

Or NormalizationDEM

Phase Unwrap Unit Circle Phase Projection

DEM

INDREX

ThisWork

© 2006 Intermap Technologies, Inc. All rights reserved.

The ‘Ideal’ Test Area(for our purpose)

Contains:Patches of alternating bare and forested areasFlat and sloped samplesSamples of varying but homogeneous canopy heights

Ground truth Air Photos (geocoded)Ground photos (georeferenced)Tree height/density/biomass samplesA truth DTM (at the 1 m level or less)

• Lidar would be great (but not available)Radar data

L- and P-Band files complete‘Good’ KzNo temporal decorrelation, good SNR

© 2006 Intermap Technologies, Inc. All rights reserved.

Location of Test Sites

© 2006 Intermap Technologies, Inc. All rights reserved.

Samboja Lestari (a compromise)

Segment 1P-Band Quad-pol

Image

© 2006 Intermap Technologies, Inc. All rights reserved.

X-Band DSMTest Area

© 2006 Intermap Technologies, Inc. All rights reserved.

The P- and L-Band Interferometric Data

P-Band Baselines nominally 15 m and 30 m

P15: KZ = - 0.026 rad / mP30: KZ = - 0.079 rad / mP-Band DEM applied directly from interferometric phase (assumes large ‘m’)

L-Band Baselines nominally 5 m and 8 m

L5: KZ = - 0.059 rad / mL8: KZ = -0.061 rad / mL-Band DEMs computed from unit circle projection of best fit among the 3 optimum complex coherences

© 2006 Intermap Technologies, Inc. All rights reserved.

Difference Surface: (X – P30) Before Correction

Test Area

meters

© 2006 Intermap Technologies, Inc. All rights reserved.

Test Area

Tree Heights 8 – 16 meters

XVVMagnitude

© 2006 Intermap Technologies, Inc. All rights reserved.

Test Area: Then (2001) and Now (2004)

IKONOS (2001)

E-SAR X-Band (2004)

© 2006 Intermap Technologies, Inc. All rights reserved.

X-Band DSM

(m)

© 2006 Intermap Technologies, Inc. All rights reserved.

P-Band DEM 30 m Baseline (P30)Kz = 0.079 rad/m

(m)

© 2006 Intermap Technologies, Inc. All rights reserved.

P-Band DEM 15 m Baseline (P15)Kz = 0.026 rad/m

(m)

© 2006 Intermap Technologies, Inc. All rights reserved.

Difference Surface (X – P30)

(m)

meters

© 2006 Intermap Technologies, Inc. All rights reserved.

Selected Profile Lines on X-Band ORI

Tree measurements

© 2006 Intermap Technologies, Inc. All rights reserved.

P30 Profile Line 01

Measured tree height: 11.9m±3.7m

P-Band (30 m)

X-Band

© 2006 Intermap Technologies, Inc. All rights reserved.

P30 Cross Profile Line 03

Measured tree height: 11.9m±3.7m

X-Band

P-Band (30m)

© 2006 Intermap Technologies, Inc. All rights reserved.

P30 Cross Profile Line 04

Measured tree height: 13.3m±2.8m

P-Band (30m)

X-Band

© 2006 Intermap Technologies, Inc. All rights reserved.

L-Band Initial Results

From unit circle projectionWhat do we expect to see?

Observed optimized coherence magnitudes

Observed points on unit circle5 and 8 m baselinesSeparation ?

© 2006 Intermap Technologies, Inc. All rights reserved.

Test Area of Interest: L-Band, Bn = 5.4 mKz = 0.073 rad/m, σ = 0.202 m-1 (0.8 db/m)

Complex Coherence for (m/I1) = 0.01, 0.03, 0.10, 0.3, 1.0, 3, 10, 30, 100

as function of canopy height

-1.0

-0.8

-0.5

-0.3

0.0

0.3

0.5

0.8

1.0

-1.0 -0.8 -0.5 -0.3 0.0 0.3 0.5 0.8 1.0

Real

Imag

inar

y

h=30 mh=20 mh=10 mh=1 m

m/I1 = 0.01

m/I1 = 1.0

© 2006 Intermap Technologies, Inc. All rights reserved.

Test Area of Interest: L-Band, Bn = 5.4 m Kz = 0.073 rad/m, σ = 0.096 m-1 (0.4 db/m)

Complex Coherence for (m/I1) = 0.01, 0.03, 0.10, 0.3, 1.0, 3, 10, 30, 100

as function of canopy height

-1.000

-0.750

-0.500

-0.250

0.000

0.250

0.500

0.750

1.000

-1.000 -0.750 -0.500 -0.250 0.000 0.250 0.500 0.750 1.000

Real

Imag

inar

y

h=30 mh=20 mh=10 mh=1 m

m/I1 = 0.01

m/I1 = 1.0

© 2006 Intermap Technologies, Inc. All rights reserved.

Test Area of Interest: L-Band, Bn = 5.4 mKz = 0.073 rad/m, σ = 0.047 m-1 (0.2 db/m)

Complex Coherence for (m/I1) = 0.01, 0.03, 0.10, 0.3, 1.0, 3, 10, 30, 100

as function of canopy height

-1.000

-0.750

-0.500

-0.250

0.000

0.250

0.500

0.750

1.000

-1.000 -0.750 -0.500 -0.250 0.000 0.250 0.500 0.750 1.000

Real

Imag

inar

y

h=30 mh=20 mh=10 mh=1 m

m/I1 = 0.01

m/I1 = 1.0

© 2006 Intermap Technologies, Inc. All rights reserved.

Test Area of Interest: L-Band, Bn = 5.4 mKz = 0.073 rad/m, σ = 0.023 m-1 (0.1 db/m)

Complex Coherence for (m/I1) = 0.01, 0.03, 0.10, 0.3, 1.0, 3, 10, 30, 100

as function of canopy height

-1.000

-0.750

-0.500

-0.250

0.000

0.250

0.500

0.750

1.000

-1.000 -0.750 -0.500 -0.250 0.000 0.250 0.500 0.750 1.000

Real

Imag

inar

y

h=30 mh=20 mh=10 mh=1 m

m/I1 = 0.01

m/I1 = 1.0

© 2006 Intermap Technologies, Inc. All rights reserved.

L-Band Coherence Mag: Opt3, 5 m Baseline, Kz = -0.06 rad/m

Bare Area

Forest

© 2006 Intermap Technologies, Inc. All rights reserved.

L-Band Coherence Mag: Opt25 m Baseline, Kz = -0.06 rad/m

Bare Area

Forest

© 2006 Intermap Technologies, Inc. All rights reserved.

L-Band Coherence Mag: Opt15 m Baseline, Kz = -0.06 rad/m

Bare Area

Forest

© 2006 Intermap Technologies, Inc. All rights reserved.

Selected Locations for Complex Coherence Plotting

© 2006 Intermap Technologies, Inc. All rights reserved.

Observed Complex Coherence L-Band 8m baseline, Test Point 10

Kz = 0.06 rad/m

Pi/2 = 26 m

Note: HV phase above HH, below VV

© 2006 Intermap Technologies, Inc. All rights reserved.

Observed Complex Coherence L-Band 8m baseline, Test Point 9

Kz = 0.06 rad/m

Pi/2 = 26 m

Note: poor separation

© 2006 Intermap Technologies, Inc. All rights reserved.

Summary and IssuesStarted work on the extraction of ground elevation from INDREX-II P-Band and L-Band data The P-Band surfaces within the forest appear plausible when normalized locally to X-Band DSM.

Additional ground truth (particularly lidar) is needed for validation

Need to resolve preliminary L-Band coherence issues that are contrary to expectation and to other INDREX inversion results

Poor phase separation in many cases VH bracketed by HH and VVUnlikely that the effect is caused by SNR Possibly temporal de-correlation but unlikely

top related