Visual Information Solutions

Report
Performing Quantitative Analysis
with Remotely Sensed Imagery in
ENVI
…we will begin shortly
The information contained in this document pertains to software products and
services that are subject to the controls of the Export Administration Regulations
(EAR). The recipient is responsible for ensuring compliance to all applicable U.S.
Export Control laws and regulations.
Performing Quantitative Analysis with
Remotely Sensed Imagery in ENVI
• The phone lines will be muted for sound quality.
• Please direct questions to the Chat window. My colleague will
be available to answer any questions.
• The presentation will be recorded and posted to the ITTVIS
website
Visual Information Solutions
Survey!
• Do you work on a Windows, Mac, or Unix machine?
Visual Information Solutions
Topics to Cover
• Concepts in Remote Sensing
• Why calibration and atmospheric correction are important data
pre-processing tasks
• Basic tools in ENVI to account for general atmospheric effects
• Advanced tools in ENVI for robust atmospheric correction
• The difference between raw data, radiance, and reflectance
data
• Applications that rely on atmospherically corrected and
calibrated data
Visual Information Solutions
Sun - Sensor Pathway
Sensor
Path Radiance
(scattered light)
Solar Irradiance
Absorbed by
atmospheric gases
Radiance (reflected and emitted energy)
absorbed
Visual Information Solutions
Solar Spectrum
0.25
0.20
Blackbody at 5900K
Solar irradiance outside atmosphere
Solar irradiance at sea level
0.15
O3
H 2O
02, H2O
H2O
H2O
H2O
H2O
0.10
0.05
0.0
H2O, CO2
H2O, CO2
0.2 0.4
0.6 0.8 1.0 1.2
H2O, CO2
1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0
3.2
Wavelength (mm)
(From Valley, 1965)
Visual Information Solutions
The Electromagnetic Spectrum
NEARVISIBLE INFRARED
(VIS)
(NIR)
SHORT WAVE
INFRARED
(SWIR)
Atmospheric Transmittance
1.0
0.9
0.8
0.7
0.6
O2
O3 O2
H2
O
CO2
H2O
CH4
O2
H 2O
H2O
0.5
H2O
0.4
H2O
CO2
0.3
H2O
0.2
0.1
0.0
400.0
700.0
1000.0
1300.0 1600.0 1900.0
Wavelength (nm)
VISIBLE
GAMMA
UV
Visual Information Solutions
INFRARED
MICROWAVE
2200.0
2500.0
RADIO
Hyperspectral and Multispectral Band Passes
Visual Information Solutions
Atmospheric Scattering as a Function of Wavelength
10
9
Relative Scatter
8
7
6
Range of
Atmospheric
Scattering
5
4
3
2
1
0.4
0.5
0.6
0.7
0.8
Wavelength (mm)
Visual Information Solutions
0.9
1.0
Raw to Radiance (Data Calibration)
• Raw DN includes:
Surface Reflectance,
Solar irradiance curve,
Atmospheric effects (scattering,
absorption),
Variation in illumination due to
topography,
Instrument response
• Raw to Radiance – remove
instrument effects
• Instrument calibration required
to derive radiance coefficients
• Raw DN * coefficients =
Radiance
Visual Information Solutions
irradiance
Atmospheric
Effects and
Surface
Reflectance on
Radiance
wavelength
irradiance
Radiance
wavelength
Atmosphere
irradiance
Radiance with atmospheric effects
wavelength
Atmosphere
Visual Information Solutions
with atmospheric effects and ground reflectance
Radiance to Reflectance
L0(l)=Lsun(l) T(l) R(l) cos(q) + Lpath(l)
•
•
•
•
•
•
L0(l) = observed radiance at sensor
Lsun(l) = Solar irradiance above atmosphere
T(l) = total atmospheric transmittance
R(l) = surface reflectance
q = incidence angle
Lpath(l) = path scattered radiance
Conversion methods generally result in “apparent reflectance”
because of topographic slope and aspect effects – variations in
illumination are not corrected
Reflectance data scaled to get to integer data (typically x 10,000)
Visual Information Solutions
The Importance of Calibration and Atmospheric
Correction
• To compare multi-date images – some data sets even
have different atmospheric properties across a scene
• To compare data sets from different sensors
• Needed for quantitative analysis, e.g., working with field
data
• convert to physical units – Radiance units: watts/sr*cm2*nm
• When using band ratios such as vegetation indices
• Reflectance data needed to compare data spectra with
library reflectance spectra – helps in identifying materials
based on their absorption features
• Or to use spectral library to map materials, image must be
in reflectance.
Visual Information Solutions
Advantages of Reflectance Data
•
Spectral features much more
apparent in reflectance data than
radiance
•
The shapes of spectra are principally
influenced by the chemical and
physical properties of surface
materials
•
Reflectance data may be analyzed
using spectroscopic methods that
isolate absorption features and relate
them to chemical bonds and physical
properties of materials.
Visual Information Solutions
Survey!
• Do you work with raw, radiance, or reflectance data?
• HSI or MSI data?
Visual Information Solutions
Multispectral Data Calibration
Visual Information Solutions
Multispectral Data Preprocessing
raw
Visual Information Solutions
radiance
Multispectral Data Preprocessing
reflectance
(with scattered light)
Visual Information Solutions
reflectance
Hyperspectral Data Preprocessing
• Data from Santa Barbara, CA
• With radiance data, the overall
shape of this spectrum is strongly
a function of the solar irradiance
spectrum and absorption by
atmospheric gases, especially
water vapor
Visual Information Solutions
water vapor
Conversion to Reflectance Methods
1. Scene-derived corrections – in-scene statistics are used
•
•
•
•
Internal Average Relative Reflectance (IAR)
Flat Field
Log Residuals
Quick Atmospheric Correction (QuAC)
2. Ground-calibration methods
•
Empirical Line
3. Radiative transfer models
•
FLAASH
Visual Information Solutions
Internal Average Relative Reflectance (IAR)
• The Internal Average
Reflectance (IAR) approach uses
the mean radiance of all the
pixels in the image as a
correction factor.
• The individual radiance values in
each pixel are divided by this
mean radiance to estimate
reflectance.
• Removes common things
• However, introduces artifacts
Visual Information Solutions
Flat Field
• Flat field - large, bright, homogenous
target
• The individual radiance values in
each pixel are divided by the mean
radiance of the flat field
• Removes things in common
Visual Information Solutions
Log Residuals
• Designed to remove solar irradiance curve, atmospheric
transmittance, instrument gain, topographic effects, and albedo
effects from radiance data
• Defined as the input spectrum divided by the spectral
geometric mean, then divided by the spatial geometric mean,
creating a pseudo reflectance image
• First calculate the spectral and spatial geometric means. Geometric
means are calculated using logarithms of the data values and are
used because the transmittance and other effects are multiplicative.
• The spectral mean is the mean of all bands for each pixel and
removes topographic effects
• The spatial mean is the mean of all pixels for each band and
accounts for the solar irradiance, atmospheric transmittance and
instrument gain
• Each image data value is then divided first by the spectral and then
by the spatial mean
Visual Information Solutions
QuAC
• QUick Atmospheric Correction is fast
• The approach is based on the finding
that the spectral standard deviation
(or endmember mean spectrum) of a
collection of endmember spectra in a
scene, is essentially spectrally flat
• Works even when the sensor was
not properly calibrated, or when the
solar illumination intensity is
unknown
• Multi- or hyperspectral data can be
raw, radiance, or apparent
reflectance
Visual Information Solutions
QUAC continued
• Does not work well in a scene that is not spectrally diverse.
The scene should have several different materials.
• The scene should have dark materials or shadows
• QuAC is Batchable
Visual Information Solutions
Empirical Line
dark target
Image radiance
Channel x
Slope=gain
bright target
dark target
Intercept=offset
bright, homogenous
target
Ground reflectance
Reflectance=gain x radiance + offset
If only one spectrum is used, then the
regression line will pass through the
origin
Visual Information Solutions
Advanced Tools in ENVI for Conversion to
Reflectance
• Radiative Transfer – based
• Models are developed that describe the radiative transfer of sunlight
in its physical interaction with the gases and particles in the
atmosphere, its interaction with the surface, and its transmission
along a different path upward through the atmosphere to the sensor.
• These models describe the solar irradiance curve, the absorption and
scattering by atmospheric gases, and the reflectance from surface
materials, all as a function of wavelength of electromagnetic radiation
and the directional angles of the sun and sensor.
• Errors arise from inadequate definition of the solar irradiance
function, variations in the illumination, imperfect models that describe
absorption by atmospheric gases, and any mis-calibration of the
sensor.
Visual Information Solutions
FLAASH
• FLAASH 4.1 – Fast Line of Site Atmospheric Analysis of Spectral
Hypercubes
• Supports many hyperspectral and multispectral Instruments: AVIRIS, HYDICE, HyMap,
Probe-1, CASI, AISA, and HYPERION, Landsat, SPOT, IRS, IKONOS, QuickBird,
ASTER, WorldView 2, etc.
• Incorporates MODTRAN4 radiative transfer code
• First, the optical characteristics of the atmosphere are estimated by using theoretical
models.
• Then, various quantities related to the atmospheric correction are computed by the
radiative transfer algorithms given the atmospheric optical properties. Then, the data
can be corrected by inversion procedures that derive the surface reflectance. Handles
clouds, cirrus and opaque.
• Gases corrected for: water vapor, ozone, oxygen, carbon monoxide,
carbon dioxide, methane, and nitrous oxide
• Water vapor the most variable and most important
• Water modeled using three-band ratios around either the1135, 940 or 820 nm
absorptions. Correction only possible where band positioning is appropriate.
• Assumes that the surface is horizontal and has a Lambertian (diffuse)
reflectance
Visual Information Solutions
Visual Information Solutions
FLAASH Parameters
• Sensor Type
• Band passes and pixel size
• Atmospheric Model - select appropriate model
• Water Retrieval – to solve radiative transfer equations, water column needed
• 1135 nm is default – use unless there are materials in scene with absorptions
at that wavelength then use 940 nm or 820 nm absorption
• Aerosol Model – not critical if visibility over 40 km
• Initial visibility – Clear: 40 to 100 km, Moderate Haze: 20—30 km, Thick Haze:
15 km or less
• Spectral Polishing – well-behaved spectra used for calculation of gain
factor. Used to remove artifacts due to:
• Errors in radiative transfer models/calculations
• Low signal in certain portions of the spectrum
• Mis-calibration of sensor
• Wavelength Recalibration – actual band positions determined from
atmospheric features. Data sets can be re-run with new wavelength file.
Visual Information Solutions
EFFORT Polishing – stand alone routine
• Empirical Flat Field Optimal Reflectance Transformation
• Bootstrapped solution – “well behaved” flat reflectance sample
spectra selected with replacement from all spectra
• bootstrapping – sampling with replacement such that selected set
can be treated as the entire population
• Statistically mild gain (close to 1) and offset (close to 0) are calculated
for each band – similar to empirical line correction
• End result – artifacts removed and spectra more of a true indication of
sensor SNR. Spectra can be more accurately compared to spectral
library spectra
Visual Information Solutions
Raw Data versus Calibrated/Corrected Results
input data
color infrared
raw data
radiance data
Maximum likelihood Maximum likelihood
classification
classification
Visual Information Solutions
Raw Data versus Calibrated/Corrected Results
input data
color infrared
raw data
radiance data
Maximum likelihood Maximum likelihood
classification
classification
Visual Information Solutions
Raw Data versus Calibrated/Corrected Results
input data
color infrared
reflectance data
NDVI
Visual Information Solutions
Dark-corrected
reflectance data
NDVI
Hyperspectral Radiance versus Reflectance Data
reflectance
radiance
input data
Visual Information Solutions
Hyperspectral SAM Results Comparison
input data
color infrared
radiance data
SAM result
reflectance data
SAM result
false positives
false positives
Visual Information Solutions
Some Applications that Rely on
Atmospherically Corrected and Calibrated Data
• Vegetation studies
• NDVI, pigments, lignin and cellulose, species and community mapping,
• Geological studies
• Mineralogy, soils, rock types
• Coastal and inland waters
• Chlorophyll, suspended sediments, bottom composition
• Snow and ice
• Snow cover fraction, grain size
• Environmental
• Oil spills, other contaminants
• Man-made infrastructure
Visual Information Solutions
For more information about ENVI’s capabilities or to
request an evaluation:
[email protected]
303.786.9900
www.ittvis.com
For upcoming seminars and training, please visit:
www.ittvis.com/EventsTraining
Visual Information Solutions

similar documents