Remote sensing –Beyond images
Mexico 14-15 December 2013
The workshop was organized by CIMMYT Global Conservation Agriculture Program (GCAP) and funded by the Bill & Melinda Gates Foundation (BMGF), the Mexican Secretariat of Agriculture, Livestock, Rural Development, Fisheries and Food (SAGARPA), the International Maize and Wheat Improvement Center (CIMMYT), CGIAR Research Program on Maize, the Cereal System Initiative for South Asia (CSISA) and the Sustainable Modernization of the Traditional Agriculture (MasAgro)
6. Satellite Spectral Comparison
MODIS
(250+ m)
ASTER
2 (R)
(15/30 m)
4 (NIR) backward
(1.2 / 3.7)
Coastal
CAVIS
(30 m)
300
3 (NIR) nadir
Red
Blue Yellow Edge
Green Red
Aerosol 1
Desert
Cloud
7 (SWIR 3)
9 (SWIR 5)
Stereo
1 (G)
WV-3
5 (SWIR 1)
6 (SWIR 2)
NIR 2
SWIR 2
NIR 1
Aerosol 2
SWIR 1
Water 2
SWIR 4
SWIR 3
NDVI - SWIR
Snow
10 (SWIR 6)
8
(SWIR4)
SWIR 6
SWIR 5
SWIR 8
SWIR 7
Aerosol 3
Cloud Height
Parallax
Green
500
Wavelength in nanometers
Water 1
700
Water 3
900
Cirrus
Aerosol 3
1100 1300 1500 1700 1900 2100 2300 2500
DigitalGlobe Proprietary and Business Confidential
6
7. CAMP processor enables high quality information &
insight extraction, and multi-temporal analysis-1
Color image of a typical 21km x 17km scene
Addressed Effects
• Opaque clouds
• Cirrus clouds
• Aerosols
• Water vapor
• Ice/Snow
• Shadows
• BRDF
Core Products
• Reflectance imagery
accurate to within 1%
absolute
• Utility masks located to
~1m CE90/LE90
4 km
visibility
11 km
7
8. CAMP processor enables high quality information &
insight extraction, and multi-temporal analysis-2
Aerosol variability on a typical 21km x 17km scene
Addressed Effects
• Opaque clouds
• Cirrus clouds
• Aerosols
• Water vapor
• Ice/Snow
• Shadows
• BRDF
Core Products
• Reflectance imagery
accurate to within 1%
absolute
• Utility masks located to
~1m CE90/LE90
4 km
visibility
11 km
8
9. CAMP processor enables high quality information &
insight extraction, and multi-temporal analysis-3
Atmospherically corrected color image of a typical 21km x 17km scene
Addressed Effects
• Opaque clouds
• Cirrus clouds
• Aerosols
• Water vapor
• Ice/Snow
• Shadows
• BRDF
Core Products
• Reflectance imagery
accurate to within 1%
absolute
• Utility masks located to
~1m CE90/LE90
4 km
visibility
11 km
9
10. Where are we headed: Enable Automated
Information Extraction
Water vapor
Distortion
Reflectance or Radiance
Enables
developing
spectral models
for automated
information
extraction
Reflectance
Aerosol
Distortion
Wavelength (nm)
Blue
Green
Red
Infrared
DigitalGlobe Proprietary and Business Confidential
10
11. NDVI values from TOA and surface reflectance
Water Vapor absorption
0.887
(~10%)
0.894
(~13%)
0.811
0.789
Scattering
11
12. NDVI from TOA reflectance
Longmont (August 10, 2011 – WV2)
NDVI from TOA reflectance
no vegetation
vigorous vegetation
12
14. Time-Series Data Set
The data set used is composed of 21 images
acquired between 2002 and 2009 by
QuickBird over the city of Denver, Colorado.
The time-series covers part of the downtown
area and includes single family houses,
skyscrapers, apartment complexes, industrial
buildings, roads/highways, urban parks, and
bodies of water.
14
17. WV2: DG-AComp accuracy
0.50
Field measurement
DG-AComp
RMSE (MS)
concrete
concrete
0.0070
asphalt
0.45
0.0127
0.40
Reflectance
0.35
0.30
0.25
0.20
0.15
RMSE (PAN)
0.10
concrete
asphalt
asphalt
0.05
0.0013
0.0018
0.00
C
B
G
Y
R
RE
N1
N2
PAN
17
18. Calibrated Tarps
The tarps were manufactured to guarantee a flat spectral response between
400 nm and 1050 nm, with a peek-to-peek variation in reflectance less than
10% between 10° and 60° off-nadir [3][4].
[3] M. Pagnutti, K. Holekamp, R.E. Ryan, R.D. Vaughan, J.A. Russell, D. Prados, and T. Stanley, “Atmospheric
correction of high spatial resolution commercial satellite imagery products using modis atmospheric products”,
in Analysis of Multi-Temporal Remote Sensing Images, May 2005, pp. 115 – 119.
[4] K. Holekamp, “NASA radiometric characterization”, in High spatial resolution commercial imagery workshop,
Reston, VA, Nov. 2004
18
19. Data Set
Location
NASA Stennis Space Center, MS
NASA Stennis Space Center, MS
NASA Stennis Space Center, MS
NASA Stennis Space Center, MS
Brookings, SD
Brookings, SD
Brookings, SD
Brookings, SD
Wiggins, MS
Wiggins, MS
Wiggins, MS
Park Falls, WI
Date
7-Feb-06
10-Jan-04
14-Nov-02
12-Mar-05
15-Sep-03
7-Sep-02
20-Jul-02
18-Oct-05
15-Mar-06
7-Jan-06
25-Jan-06
5-Aug-05
Sat. Az.
208.2
259.7
274.2
268.4
284.4
191.4
349.7
298.2
319.5
300.0
296.6
261.5
Sat. El.
83.2
88.2
79.2
77.7
83.1
74.6
64.3
73.2
76.7
68.0
69.3
69.4
This set of 12 QuickBird images includes four 20 m2 spectrally-flat tarps having
nominally 3.5, 22, 34, and 52% reflectance in the visible through NIR spectral
region.
19
20. QB: DG-AComp accuracy
(52, 34, 22, and 3.5% reflectance)
1.0
DG-AComp
RMSE (MS)
52%
0.8
0.0155
34%
0.0127
22%
0.0155
3.5%
0.9
0.0082
Reflectance
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
B
G
R
NIR
20
21. Information Extraction: Exploiting various
dimensions of imagery
Spectral
Spatial/
Morphological
Temporal
DigitalGlobe Proprietary and Business Confidential
21
22. Information Extraction: Model Portability
Istanbul – WV02: Feb. 19, 2010
Honolulu – WV02: Apr. 25, 2010
Rio de Janeiro – WV02: Jan. 19, 2010
New York – WV02: Dec. 18, 2009
Six classes of interest:
1. Grass
2. Tree
3. Water
4. Soil
5. Built-up
6. Shadow
22
23. Information Extraction: Model Portability
Three different experiments using training pixels from:
1. Honolulu
2. Honolulu and Istanbul
3. Honolulu, Istanbul, and New York
DN
In all cases, the image of Rio de Janeiro has been used
ONLY for validation.
DN
Kappa Coefficient
Surf. Refl.
Surf. Refl.
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
Honolulu
Honolulu, Istanbul
Honolulu, Istanbul,
NewYork
23