ACCES au laboratoire GIPSA-lab                                             Feuille de routeX.	  Ceamanos.	  26/07/11	     ...
Mars	  observed	  by	  Viking	  Orbiters	  X.	  Ceamanos.	  26/07/11	     IGARSS	  2011	  -­‐	  xavier.ceamanos@obs.ujf-­‐...
Mars	  observed	  by	  Viking	  Orbiters	                                                                                 ...
Mars	  observed	  by	  Viking	  Orbiters	                                                                                 ...
X.	  Ceamanos.	  26/07/11	     IGARSS	  2011	  -­‐	  xavier.ceamanos@obs.ujf-­‐grenoble.fr	     5	  
HiRISE@MRO	  (snapshot	                    HiRISE@MRO	  25	  cm/pix)	                                                     ...
HiRISE@MRO	  (snapshot	                    HiRISE@MRO	  25	  cm/pix)	                                                     ...
HiRISE@MRO	  (25	  cm/pix)	                              CRISM	  pixel	  	                               footprint	  Mars	...
X.	  Ceamanos.	  26/07/11	     IGARSS	  2011	  -­‐	  xavier.ceamanos@obs.ujf-­‐grenoble.fr	     9	  
Raw	  image	            Artifact	  cleaning	       Photometric	  correction	       Atmospheric	  correction	             C...
X.	  Ceamanos.	  26/07/11	     IGARSS	  2011	  -­‐	  xavier.ceamanos@obs.ujf-­‐grenoble.fr	     11	  
BPSS endmember spectra                                                          BPSS Dark source                          ...
X.	  Ceamanos.	  26/07/11	     IGARSS	  2011	  -­‐	  xavier.ceamanos@obs.ujf-­‐grenoble.fr	     13	  
VCA	                       MVC-­‐NMF	                        spatial-­‐VCA	            BPSS	                              ...
VCA	                       MVC-­‐NMF	                        spatial-­‐VCA	            BPSS	                              ...
VCA	                       MVC-­‐NMF	                        spatial-­‐VCA	            BPSS	                              ...
VCA	                       MVC-­‐NMF	                        spatial-­‐VCA	            BPSS	                              ...
X.	  Ceamanos.	  26/07/11	     IGARSS	  2011	  -­‐	  xavier.ceamanos@obs.ujf-­‐grenoble.fr	     18	  
X.	  Ceamanos.	  26/07/11	     IGARSS	  2011	  -­‐	  xavier.ceamanos@obs.ujf-­‐grenoble.fr	     19	  
0.6 Apparent reflectance   0.4                                                                                       0.5  ...
Sources:	  Dark,	  strong	  bright,	  weak	  bright	                                                                      ...
Sources:	  Dark,	  strong	  bright,	  weak	  bright	                            VCA	                                     B...
X.	  Ceamanos.	  26/07/11	     IGARSS	  2011	  -­‐	  xavier.ceamanos@obs.ujf-­‐grenoble.fr	     23	  
X.	  Ceamanos.	  26/07/11	     IGARSS	  2011	  -­‐	  xavier.ceamanos@obs.ujf-­‐grenoble.fr	     24	  
HiRISE@MRO	  (25	  cm/pix)	  X.	  Ceamanos.	  26/07/11	     IGARSS	  2011	  -­‐	  xavier.ceamanos@obs.ujf-­‐grenoble.fr	  ...
Dark	  features	                                                                                                          ...
Dark	  features	                                                                                                          ...
HiRISE PSP_002482_1255                 CRISM frt000042aa        Registration correlation coefficient                      ...
HiRISE PSP_002482_1255                    CRISM frt000042aa              Registration correlation coefficient             ...
HiRISE PSP_002482_1255                     CRISM frt000042aa              Registration correlation coefficient            ...
VCA	                                  BPSS	                                    MVC-­‐NMF	                       spatial-­‐...
X.	  Ceamanos.	  26/07/11	     IGARSS	  2011	  -­‐	  xavier.ceamanos@obs.ujf-­‐grenoble.fr	     32	  
X.	  Ceamanos.	  26/07/11	     IGARSS	  2011	  -­‐	  xavier.ceamanos@obs.ujf-­‐grenoble.fr	     33	  
ACCES au laboratoire GIPSA-lab                                             Feuille de routeX.	  Ceamanos.	  26/07/11	     ...
X.	  Ceamanos.	  26/07/11	     IGARSS	  2011	  -­‐	  xavier.ceamanos@obs.ujf-­‐grenoble.fr	     35	  
Dark	  Strong	  bright	  Weak	  bright	                                                    Non-­‐linear	  residue	        ...
VZA=-­‐30º	      Dark	  Strong	  bright	  Weak	  bright	                                                                  ...
VZA=30º	                                                                                  VZA=-­‐30º	      Dark	  Strong	 ...
Upcoming SlideShare
Loading in …5
×

BLIND SOURCE SEPARATION OF HYPERSPECTRAL DATA IN PLANETARY REMOTE SENSING: ENDMEMBER EXTRACTION AND VALIDATION

824 views

Published on

  • Be the first to comment

BLIND SOURCE SEPARATION OF HYPERSPECTRAL DATA IN PLANETARY REMOTE SENSING: ENDMEMBER EXTRACTION AND VALIDATION

  1. 1. ACCES au laboratoire GIPSA-lab Feuille de routeX.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   1   GIPSA-lab
  2. 2. Mars  observed  by  Viking  Orbiters  X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   2  
  3. 3. Mars  observed  by  Viking  Orbiters   Geographical  linear  mixture   Mineral   dust   Pixel  size  X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   3  
  4. 4. Mars  observed  by  Viking  Orbiters   Geographical  linear  mixture   Mineral   dust   Pixel  size  X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   4  
  5. 5. X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   5  
  6. 6. HiRISE@MRO  (snapshot   HiRISE@MRO  25  cm/pix)   THEMIS  ~100  m/pix   40  km   20  km  X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   6  
  7. 7. HiRISE@MRO  (snapshot   HiRISE@MRO  25  cm/pix)   THEMIS  ~100  m/pix  X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   7  
  8. 8. HiRISE@MRO  (25  cm/pix)   CRISM  pixel     footprint  Mars  Reconnaissance  Orbiter  CRISM   X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   8  
  9. 9. X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   9  
  10. 10. Raw  image   Artifact  cleaning   Photometric  correction   Atmospheric  correction   Clean  image  X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   10  
  11. 11. X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   11  
  12. 12. BPSS endmember spectra BPSS Dark source 0.25 0.5 0.4 0.2 REFF value 0.3 0.2 0.15 0.1 A 0.1 B 0 e1 e2 e3 e4 e5 e6 e5’ 0 50 100 150 200 250 0 50 100 150 200 250 0.5 BPSS Strong bright source 0.4 mn : spectral  signature  of   BPSS Weak bright source 0.4 BPSS endmember 1 0.35 endmember  n BPSS endmember 2 0.3 REFF value 0.3 0.5 0.45 0.25 0.2 0.4 0.2 0.4 0.1 C 0.35 0.15 D 0.3 0.1 e1’ e2’ e3’ e4’ e6’ 0.3 0 0.25 0.05 0 50 100 150 200 250 0 50 100 150 200 250 CRISM spectral band 0.2 CRISM spectral band 0.2 0.15 0.1 0.1 0.05 BPSS endmember 4 sn : abundance  map  of   BPSS endmember 5 0.6 endmember  n   0.8 0.5 0.7 0.4 0.6 0.5 0.3 0.4 0.2 0.3 0.2 0.1 0.1X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   12  
  13. 13. X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   13  
  14. 14. VCA   MVC-­‐NMF   spatial-­‐VCA   BPSS   [Nascimento’05]   [Miao’07]   [Zortea’09]   [Moussaoui’06]   Geometric  method   First   Minimum  volume   Incorporation  of   Statistical   with  pure  pixel   principles:   constraint   spatial  information   approach   assumption   -­‐  Fast  &  efficient   -­‐  Bayesian   -­‐  Less-­‐prevalent   -­‐  Homogeneous  Advantages:   -­‐  Endmembers  are   endmembers   endmembers   framework   physical   -­‐  Error  bars   -­‐  Non-­‐physical   -­‐  Spatially-­‐ -­‐  Impact  of  noise   spectra   -­‐  Non-­‐physical   confined  and  less-­‐Drawbacks:   -­‐  Less-­‐prevalent   spectra   prevalent   -­‐  High   endmembers   computational   endmembers   time   X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   14  
  15. 15. VCA   MVC-­‐NMF   spatial-­‐VCA   BPSS   [Nascimento’05]   [Miao’07]   [Zortea’09]   [Moussaoui’06]   Geometric  method   First   Minimum  volume   Incorporation  of   Statistical   with  pure  pixel   principles:   constraint   spatial  information   approach   assumption   -­‐  Fast  &  efficient   -­‐  Bayesian   -­‐  Less-­‐prevalent   -­‐  Homogeneous  Advantages:   -­‐  Endmembers  are   endmembers   endmembers   framework                           physical   -­‐  Error  bars   -­‐  Non-­‐physical   -­‐  Spatially-­‐ -­‐  Impact  of  noise   spectra   -­‐  Non-­‐physical   confined  and  less-­‐Drawbacks:   -­‐  Less-­‐prevalent   spectra   prevalent   -­‐  High   endmembers   computational   endmembers   time   X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   15  
  16. 16. VCA   MVC-­‐NMF   spatial-­‐VCA   BPSS   [Nascimento’05]   [Miao’07]   [Zortea’09]   [Moussaoui’06]   Geometric  method   First   Minimum  volume   Incorporation  of   Statistical   with  pure  pixel   principles:   constraint   spatial  information   approach   assumption   -­‐  Fast  &  efficient   -­‐  Bayesian   -­‐  Less-­‐prevalent   -­‐  Homogeneous  Advantages:   -­‐  Endmembers  are   endmembers   endmembers   framework                           physical   -­‐  Error  bars   -­‐  Non-­‐physical   -­‐  Spatially-­‐ -­‐  Impact  of  noise   spectra   -­‐  Non-­‐physical   confined  and  less-­‐Drawbacks:   -­‐  Less-­‐prevalent   spectra   prevalent   -­‐  High   endmembers   computational   endmembers   time   X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   16  
  17. 17. VCA   MVC-­‐NMF   spatial-­‐VCA   BPSS   [Nascimento’05]   [Miao’07]   [Zortea’09]   [Moussaoui’06]   Geometric  method   First   Minimum  volume   Incorporation  of   Statistical   with  pure  pixel   principles:   constraint   spatial  information   approach   assumption   -­‐  Fast  &  efficient   -­‐  Bayesian   -­‐  Less-­‐prevalent   -­‐  Homogeneous  Advantages:   -­‐  Endmembers  are   endmembers   endmembers   framework                           physical   -­‐  Error  bars   -­‐  Non-­‐physical   -­‐  Spatially-­‐ -­‐  Impact  of  noise   spectra   -­‐  Non-­‐physical   confined  and  less-­‐Drawbacks:   -­‐  Less-­‐prevalent   spectra   prevalent   -­‐  High   endmembers   computational   endmembers   time   X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   17  
  18. 18. X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   18  
  19. 19. X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   19  
  20. 20. 0.6 Apparent reflectance 0.4 0.5 0.4 0.3 0.3 0.2 0.2 0.1 A! B!Spectral  product:  spectral  signatures   1 2 3 4 5 6 0 1 Spatial  product:  abundance  maps   2 3 4 5 6 1.32 1.65 1.98 2.31 2.64 1.32 1.65 1.98 2.31 2.64 MVC NMF associated spectra spatial VCA associated spectra R1.1  um   B2.3  um   0.4 0.4Apparent reflectance 0.3 0.3 A! B! 0.2 0.2 0.1 0.1 C! 1.32 1 2 1.65 3 4 1.98 5 2.31 6 2.64 D! 1.32 1 2 1.65 3 4 1.98 5 2.31 6 2.64 Wavelength in microns Wavelength in microns C! D! Dark   Strong  bright   Weak  bright   Final  product:  composite   abundance  map!   X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   20  
  21. 21. Sources:  Dark,  strong  bright,  weak  bright   C! HiRISE  image   [Ceamanos  TGRS  2011]   MVC-­‐NMF  composite  map  X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   21  
  22. 22. Sources:  Dark,  strong  bright,  weak  bright   VCA   BPSS   A! B! MVC-­‐NMF   spatial-­‐VCA   C! D!X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   22  
  23. 23. X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   23  
  24. 24. X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   24  
  25. 25. HiRISE@MRO  (25  cm/pix)  X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   25  
  26. 26. Dark  features   reference   abundance   map CO2  ice   CRISM HiRISE 150 m 50 m Detail  of  the  Russell  dune  observed  by  the  CRISM  and  the  HiRISE   instruments.  CRISM  frt42aa  in  blue,  HiRISE  PSP_002482_1255  in  green  X.  Ceamanos.  26/07/11     IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   26  
  27. 27. Dark  features   reference   abundance   map A! CO2  ice   C! CRISM HiRISE 150 m 50 m Detail  of  the  Russell  dune  observed  by  the  CRISM  and  the  HiRISE   instruments.  CRISM  frt42aa  in  blue,  HiRISE  PSP_002482_1255  in  green  X.  Ceamanos.  26/07/11     IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   27  
  28. 28. HiRISE PSP_002482_1255 CRISM frt000042aa Registration correlation coefficient 1 0.25  m/pix   18  m/pix   Avg.  Corr.  =   0.9 29862×63004  pix   604×420  pix   0.7   0.8 0.7 0.6 HiRISE  image   0.5 0.4 0.3 1.  Registration   CRISM  image   0.2 0.1 A! B! C! 0 Reg.  HiRISE  image   2.  Classification   Classification  map   3.  Pixel  counting   Abundance  map   (ground  truth)  X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   28  
  29. 29. HiRISE PSP_002482_1255 CRISM frt000042aa Registration correlation coefficient 1 0.25  m/pix   18  m/pix   Avg.  Corr.  =   0.9 29862×63004  pix   604×420  pix   0.7   0.8 0.7 0.6 HiRISE  image   0.5 0.4 0.3 1.  Registration   CRISM  image   0.2 0.1 A! B! C! 0 Reg.  HiRISE  image   2.  Classification   Classification  map   Classification map Ground truth 1 A! B! 0.9 0.8 3.  Pixel  counting   0.7 0.6 0.5 Abundance  map   0.4 (ground  truth)   0.3 0.2 0.1 0X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   29  
  30. 30. HiRISE PSP_002482_1255 CRISM frt000042aa Registration correlation coefficient 1 0.25  m/pix   18  m/pix   Avg.  Corr.  =   0.9 29862×63004  pix   604×420  pix   0.7   0.8 0.7 0.6 HiRISE  image   0.5 0.4 0.3 1.  Registration   CRISM  image   0.2 0.1 A! B! C! 0 Reg.  HiRISE  image   2.  Classification   Classification  map   Classification map Ground truth 1 Pixel  counting  for  two  CRISM  pixels   A! B! 0.9 0.8 3.  Pixel  counting   a(xi)=0.10   a(xj)=0.35   0.7 0.6 0.5 Abundance  map   0.4 (ground  truth)   0.3 0.2 0.1 0X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   30  
  31. 31. VCA   BPSS   MVC-­‐NMF   spatial-­‐VCA  Ground  truth   Registration  accuracy   •  10%  error  between  ground  truth  and  unmixing  results   •  MVC-­‐NMF  obtains  the  best  r =  0.83  and  ε =  0.08   •  BPSS  provides  accurate  abundances   •  VCA  provides  underestimated  abundances   •  spatial-­‐VCA  does  not  extract  the  dark  source  satisfactorily X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   31  
  32. 32. X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   32  
  33. 33. X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   33  
  34. 34. ACCES au laboratoire GIPSA-lab Feuille de routeX.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   34   GIPSA-lab
  35. 35. X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   35  
  36. 36. Dark  Strong  bright  Weak  bright   Non-­‐linear  residue   due  to  unaccurate   atmospheric  correction   X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   36  
  37. 37. VZA=-­‐30º   Dark  Strong  bright  Weak  bright   target   Non-­‐linear  residue   due  to  unaccurate   atmospheric  correction   X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   37  
  38. 38. VZA=30º   VZA=-­‐30º   Dark  Strong  bright  Weak  bright   target   Non-­‐linear  residue   due  to  unaccurate   atmospheric  correction   X.  Ceamanos.  26/07/11   IGARSS  2011  -­‐  xavier.ceamanos@obs.ujf-­‐grenoble.fr   38  

×