Compact polarimetry SAR for natural surface characterization: an attractive compromise

Size: px

Start display at page:

Download "Compact polarimetry SAR for natural surface characterization: an attractive compromise"

James Holmes
6 years ago
Views:

1 Compact polarimetry AR for natural surface characterization: an attractive compromise P Dubois-Fernandez (ONERA) JC ouyris (CNE), Angelliaume (ONERA), K Raney (John Hopkins University)

2 Outline of the presentation Introduction to Compact Polarimetry ystem consideration PolAR analysis PolInAR analysis Atmospheric effect Recommendations from PolINAR 2007 Conclusions

3 Full polarimetry: the burden of the chronogram H V H V H V Transmit : dual versus single PRF PRF x 2 Average transmit power X 2 + Receive: dual versus single Data vol. Data vol. x 2 + Range ambiguity condition : wath wath / 2

4 Example with ALO system Mode wath Resolution HH 70km 10m HH/HV or VV/VH 70km 20m Quad-pol 30km 30m Incidence angle

5 Compact Polarimetry: optimisation Compact polarimetry 1 polarization on transmit 2 polarizations on receive (with phase coherence) ALO example: H on transmit, H and V on receive V on transmit, H and V on receive What is the best polarization on transmit? What are the best polarizations on receive?

6 Background publications J.C. ouyris,. Mingot, "Polarimetry based on one transmitting and two receiving polarizations : the p/4 mode", Proceedings IGAR'02, Toronto, Canada, July 2002 Concept POLINAR 2003 J.C. ouyris, P. Imbo, R. Fjortoft,. Mingot, J.. Lee, "Compact polarimetry based on symmetry properties of geophysical media : the p/4 mode", IEEE Transactions on Geoscience and Remote ensing, vol. 43, n 3, Mars 2005, pp N. tacy, M. Preiss, "Compact polarimetric analysis of X band AR data", in. Proc. EUAR'06, Dresden, Germany, May 2006 Keith Raney, "Hybrid polarimetric AR Architecture", Proceedings IGAR'06, July 2006 Keith Raney, "Dual-Polarized AR and tokes Parameters," IEEE Geoscience and Remote ensing Letters, vol. 3, pp , POLINAR2007 ouyris, Raney, Ainswoth, Lardeux, Dubois-Fernandez

7 The revenge of deserted symmetry Current approach in dual-pol V reception transmission ALO, ENVIAT H H Or transmission V V H dual case reception Reception of two «unbalanced signals» on Co- and X- channels

8 Why is it a poor choice for natural targets? Acquisition of HH and HV Co-pol and cross-pol are not correlated Reflection symmetry hypothesis s HH s HV Phase information is useless s HV Phase and correlation between HH and VV : important parameters s VV 0

9 V reception transmission H+V 2 H reception

10 V transmission H+j.V 2 reception H reception

11 Compact polarimetry architecture V V H H Antenna V H 90 / 90o -90 or 0 o H LNA V LNA Transmitter & waveform Timing and control H Rx channel V Rx channel k 1 k 2 Circular transmit: 90 H+V transmit: 0

12 Three modes for evaluation Name PC1 PC2 PC3 Mode 45 Circular Hybrid Transmit 45 RC RC Receive 1 H RC H Target Vector: Receive 2 V LC V HH HV k r + = HV + VV k k 1 2 HH + 2 jhv HH + VV HH HV ± ± jhv jvv VV

13 Equivalence of PC2 et PC3 r HH + 2 jhv VV k PC 2 = HH + VV r k = 1 1 r k PC3 j r j = PC 2 k PC 3 HH HV We will consider only PC3 from now on. + + jhv jvv

14 PolAR Analysis

15 POLAR analysis with CP data? Learn to use it as is: K Raney and the new tokes parameters Cloude and the dual-pol entropy alpha parameters Reconstruct a pseudo full polar information JC ouyris T Ainsworth C Lardeux

16 Raney s approach: tokes parameters Key to the design tokes parameters are necessary and sufficient to to fully characterize the received EM field tokes parameters require relative phase Transmit circular polarization NB: NB: NOT NOT quad-pol; backscattered field field only, only, not not the the scattering matrix matrix Hybrid Polarity Conventional 1 = < E H 2 + E V 2 > < E L 2 + E R 2 > 2 = < E H 2 - E V 2 > 2 Re < E L E R > 3 = 2 Re < E H E V > 2 Im < E L E R > 4 = - 2 Im < E H E V > - < E L 2 - E R 2 >

17 tokes child parameters (selected) Degree of polarization m = ( ) ½ / 1 Degree of linear polarization m L = ( ) ½ / 1 Circular polarization ratio µ C = ( 1-4 ) / ( ) Relative phase δ = arctan ( 4 / 3 ) Fundamental; 1:1 mapping wrt Entropy E E ~ (1 m 2 ) γ, γ~0.74 Indicator of volume vs. subsurface scattering if m L > 0 (if CP transmission) Indicator of scattering associated with planetary ice deposits or dihedrals: µ C > ~0.4 ensitive indicator of double bounce backscattering

RH-Pol bacscatter Decomposition in m- δ space Relative H-V phase δ (degrees) +180 90 0-90 -180 0 Degree of polarization m tokes-parameter based

transmissions, indicating single-bounce (specular) or triple-bounce backscatter Phase signal of same-sense circularlypolarized return in response to

18 RH-Pol bacscatter Decomposition in m- δ space Relative H-V phase δ (degrees) Degree of polarization m tokes-parameter based transformation from hybrid-polarity AR image data to m-δ feature space 1 Expected opposite-sense phase signal in response to circularly polarized transmissions, indicating single-bounce (specular) or triple-bounce backscatter Phase signal of same-sense circularlypolarized return in response to circularly polarized transmissions, which indicates double bounce (dihedral) or coherent backscatter effect (volumetric water ice) Polarimetric AR data, Courtesy, JPL

19 ouyris approach: Pseudo covariance matrix k k 1 1 k k 1 2 k k 2 2 k k 4 measures 1 2 HH HV VV HH HH HH HH HV VV HV HV HV 9 unknowns HH HV VV VV VV VV

20 Reconstruction of the covariance matrix H1: Reflection symmetry k k 1 1 k k 1 2 k k 2 2 k k 4 measures 1 2 s HH s HV s VV s HH VV 0 HV HH HH 0 HV 0 HV 0 5 unknowns HH VV 0 VV VV

21 Reconstruction of the covariance matrix H2: HV linked to the randomness k k 1 1 k k 1 2 s HH k k 4 s 2 2 k k s 1 2 HV HH s + HV s VV s VV = (1 4 measures 5 unknowns + 1 equation s HH VV HH 0 s s HH HH HH HH s VV s VV HV s 0 0 VV HV ) HH VV 0 VV VV

22 Reconstruction of the covariance matrix Examples with X, L and P bands Modes PC1 and PC3 Name Mode Transmit Receive 1 PC H PC2 Circular RC RC PC3 Hybrid RC H Receive 2 V LC V

23 P Band over Nezer Forest

24 P-Band: Reconstruction of the HH term Mode π/4 Mode hybrid Window 7x7, 5 iterations, scale =[-20dB,+20dB]

25 P-Band: Reconstruction of the Hv Mode π/4 Mode hybrid Window 7x7, 5 iterations, scale =[-20dB,+20dB]

26 P-Band: Reconstruction of the VV Mode π/4 Mode hybrid Window 7x7, 5 iterations, scale =[-20dB,+20dB]

27 P-Band: Reconstruction Hh-Vv coherence Mode π/4 Mode hybrid Window 7x7, 5 iterations, scale =[0,1]

28 P-Band: Reconstruction Hh-Vv phase Mode π/4 Mode hybrid Window 7x7, 5 iterations, scale =[0,360 ]

29 P-Band: False color representation Full polar π /4 Hybrid R: HH, G=HV, B=VV

30 L Band: Reconstruction performance HH VV HV HH reconstructed VV recons. HV recons. L band HH/VV coherence HH/VV recons. π/4 mode Full pol

31 X-Band: False color representation Full polar π /4 Hybrid R: HH, G=HV, B=VV

32 X Band: Reconstruction HH Mode π/4 Mode hybrid Window 7x7, 5 iterations, scale =[-40dB,0dB]

33 X Band: Reconstruction Hv Mode π/4 Mode hybrid Window 7x7, 5 iterations, scale =[-40dB,0dB]

34 X Band: Reconstruction Vv Mode π/4 Mode hybrid Window 7x7, 5 iterations, scale =[-40dB,0dB]

35 X-Band: Reconstruction Hh-Vv coherence Mode π/4 Mode hybrid Window 7x7, 5 iterations, scale =[0,1]

36 X Band: Reconstruction Hh-Vv phase Mode π/4 Mode hybrid Window 7x7, 5 iterations, scale =[0,360 ]

37 Compact PolInAR analysis RVOG model P-Band specificity Inversion results from Compact Polarimetry

38 PolInar data for biomass Polarimetry: identifies the scattering type ingle-bounce Double-bounce Diffuse Interferometry: elevation PolInar: associates elevation to scattering type Hv Vv Hh Model Biophysical parameters tree height tree biomass attenuation

39 PolInar analysis HH, HV, VV, HH+VV, HH-VV

40 Inversion PolInAR tandard Inversion RVoG Adapted to P-Band Known extinction Range of extinction Time-frequency γ g

41 PolInAR results at P band σ x =0.3 db/m

42 PolInAR results at P band σ x =0.3 db/m

43 PolInAR results at P band Tree height [m] σ x =0.3dB/m Estimated tree height [m] RM error = 1.2m

44 PolInAR inversion at P band Tree height [m] Att =0.1 Att=0.5 Att=0.3 Estimated tree height [m]

45 Compact PolInar For all ψ, χ, compute k k 2 M γ 1 1 k1 γ 2 k 2 1 γ γ 2 k( ψ, χ) = cosψ. k1 + sinψ. k2e γ ψχ Inversion as before jχ

46 Compact polinsar inversion FP PC1 PC Measured Height [m] Estimated Height [m] PC1: 45,(H,V) PC3: RC, (H,V)

47 Compact PolInar inversion PC1 PC2 PC Quad Pol sector [ ] Compacl pol sector [ ] PC1: 45,(H,V) PC2: RC,(RC,LC) PC3: RC, (H,V)

48 Compact Polarimetry and atmospheric effects Introduction Mode selection PolAR PoLinAR

49 Ionospheric Effects Faraday rotation cos Ω M = sin Ω Dispersivity cintillation sin Ω cos Ω cos Ω sin Ω sin Ω cos Ω What is the spatial scale of the TEC variation? HH HV HV VV

50 Ionospheric Correction on Full Polar data M RR = RR M LL = LL Arg ( M M M M LR RL LR RL ) jω = e 2 = e 2 jω LR RL On any targets = 4Ω Bickel and Bates, Proc IRE, 1965.

51 Transmitting a circular polarized wave, a logical choice Circular pol Linear pol Transmitting a circularly polarized wave guaranties that the scattering element always sees the same incident wave. This is not the case with any other polarizations. The scattered field is traveling back to the radar and is rotated. On reception, two orthogonal polarization provides the full reception of the scattered field.

52 electing the compact polarimetry modes Full polarimetry is four dimensional 2 orientation and 2 ellipticity angles Compact polarimetry provides access to a sub-space of the full polar observation space 2 dimensional : one orientation and one ellipticity angles The observed sub-space varies with ionosphere, except when transmitting circular This invariance is essential to provide consistent information From now on, PC3 (Circular on transmit, H and V on receive)

53 Ionospheric Correction & Compact Polarimetry New concept and on-going work Two different problems PolAR analysis PolInAR analysis Promising preliminary results in both cases Evolving rapidly To be completed

54 PolAR and ionophere: possible approaches Estimation of Ω before and after acquisition A few full polar bursts to estimate Ω Acquisition CP and corrections. patial variability of ionosphere? Estimation of Ω on particular targets ingle bounce surfaces : Estimation: FP CP image acquisition FP Im( Im( LL LR LH LV Arg( ) M LR M RL = 180 or0 ) = 90 2Ω = Arg ( M M LL LR ) 180 ) = 4Ω

55 Compact PolAR and ionosphere -90 Amplitude Phase LH LV Coherence LH LV

56 PolAR and ionosphere Arg(LH LV)= -90 on bare surface (high coherence) varies with ionosphere Arg(LL LC)=180 on bare surface (high coherence) shifts by 2 Ω with ionosphere Potential techniques to estimate the ionosphere on compact polarimetry acquisition. How accurately? Investigation still going on Once the ionosphere is estimated, it is possible to invert it (in the circular transmit case)

57 Faraday rotation and compact PolInar Mode PC3 k r ref HH HV ± ± jhv jvv Ω 1 r k Ω int 2 r k ref Ω 1 HH(cos HH( j Different angles loss of coherence int k r Ω 2 2 Ω + j sin ΩcosΩ) + jhv + VV ( sin Ω + j sin ΩcosΩ) 2 2 sin Ω + sin ΩcosΩ) + HV + VV ( j cos Ω sin ΩcosΩ) 2 But we are in the same observation sub-space

58 Faraday rotation and compact PolInar k r ref Ω 1 r k Ω int 2 r k ref Ω 1 int k r Ω 2 k k 1 2 Rotation α k k 1 1 cosα k sin α + k 2 2 sin α cosα γ 1( ( α ) γ α ) 2

59 Compact PolInAR: Estimation of Faraday (1/2) γ ( α ) 1 γ ( α ) 1 Ω = 0 1 Ω = 20 2 Different areas: ionospheric differences of 20

60 Compact PolInAR: Estimation of Faraday (2/2) 180 γ ( α ) 1 γ ( α ) 1 Ω = 0 1 Ω = 20 2 Different areas: ionospheric differences of 20

61 Compact PolInAR Two solutions (±1) provide the same inversion The line is just rotated by 180 Given two acquisitions acquired with a different ionophere in mode PC3 (CH,CV), we can correct for the differential ionospheric angle to within a few degrees How robust is the inversion to a few degree differential ionosphere?

62 Robustness of inversion r k Ω r k Ω ref 1 ref ( ) 1 2 γ Ω,Ω 1 r r k Ω k int 2 Ω int ( ) 1 2 γ h Ω,Ω 2 Different ionosphere on both acquisitions

63 5 4,5 4 3,5 3 2,5 2 PC1 PC3 1,5 1 0, Differential ionospheric angle [ ] PC1: 45,(H,V) PC3: RC, (H,V) Different ionosphere on both acquisitions RM height error [m]

64 Compact PolInAR inversion Estimate the differential Faraday angle by maximizing the coherences over several areas. Correct one acquisition with this estimation Compute a large set of coherence by varying the orientation and ellipticity on the receive antenna Compute the line elect the ground point Intersection of the line with the curve of the volume only coherence associated with a given attenuation

65 Compact polarimetry What do we know? What else do we need to know?

66 Conclusions from on-going work (1/2) Complementary to full polarimetry Compatible system design Provides better coverage hould be considered as another mode of operation, like dual-pol mode for ALO PolAR analysis pecific compact polarimetric analysis possible (Raney, Cloude) Reconstruction of the full polar covariance matrix for natural targets Reconstruction possible for urban targets (Ainsworth) Information at pixel level is not preserved Compact PolInAR Very good performance for RVOG inversion at P Band Analysis to be extended to other datasets, different trees, mixed forest, significant slopes

67 Conclusions from on-going work (2/2) Circular transmit provides invariance of the observation space with respect of ionosphere PolAR and ionosphere Potential schemes for correction for PolAR data promising but to be investigated further PolInAR and ionosphere PolInAR ionospheric correction has been demonstrated in a test case. mall improvements to the inversion are necessary.

68 Recommendations for POLINAR2007 (1/2) Compact polarimetry session chaired by Freeman and ouyris The effect of terrain slope on CP signals must be carefully assessed Faraday effects : it seems that some CP options (circular pol. transmission) would tolerate Faraday rotation to up to Issue of 1st concern as P band is a good candidate for CP Comparison between CP and FP must include both POLAR & POL-INAR aspects Antenna trade-offs must be investigated Technologically speaking, the design of a quad-pol mode and of a CP mode are compatible : CP can be easily integrated on FP architectures

69 Recommendations for POLINAR2007 (2/2) Applications Earth observation : working point should be identified in the framework of biomass vegetation applications (P band) assumptions for FP reconstruction usually verified Ice applications : geophysical parameters not known precisely. Relations with POL-INAR parameters to be established first urface-ubsurface large incidence angles appear to be compliant with CP : need more investigation

70 More specifically Calibration scheme Influence of the accuracy of the dephasing device Importance of the accurate balancing of the H and V on transmit More specific work needed ystem design, internal calibration PolAR analysis Evaluation has to be done for each specific application for a representative range of conditions PolInAR analysis Evaluation has to be done for a representative range of conditions Plain interferometry is compromised by ionosphere. Better knowledge of ionosphere To be taken into account in the studies Are the models accurate in the ionospheric effect description? equencing some modes to get to the ionosphere More on the spatial variability

About calibration Degree-of-polarization / phase feature space 1 m 0 180 o Essentially a two-dimensional histogram of relative phase and degree of polarization

71 About calibration Degree-of-polarization / phase feature space 1 m o Essentially a two-dimensional histogram of relative phase and degree of polarization M-delta feature space for AirAR an Francisco 0 δ -180 o Polarization histogram 0 Degree of polarization m Phase histogram Relative 0 H-V 90 phase 180 δ

72 Phase Calibration Phase histogram This example: ~15 o ~5 o Transmit H/V phase Appears as equal shifts of the 90-degree peaks toward (or away from) zero Receive H/V phase Appears as a linear shift of both peaks toward larger (or smaller) relative phase

Faraday rotation estimation from unfocussed ALOS PALSAR raw data

Faraday rotation estimation from unfocussed ALOS PALSAR raw data arco Lavalle 1 3, E. Pottier 2, D. Solimini 1, N. iranda 3 1 DISP, Tor Vergata University, Rome, Italy 2 IETR UR CNRS 6164, University of