Goals: SUGIANTO, Ray MERTON, Shawn LAFFAN and Jon … · 2010. 11. 10. · SUGIANTO, Ray MERTON, Shawn LAFFAN and Jon HUNTINGTON 2 Objective A multi-angular and multi-temporal study

1

11

3rd FIG Regional Conference, Jakarta, Indonesia, October 3-7, 2004

A MultiA Multi--angular and Multiangular and Multi--temporal Study temporal Study using CHRIS Hyperspectral Imagery using CHRIS Hyperspectral Imagery

SUGIANTO, Ray MERTON, Shawn LAFFAN and Jon HUNTINGTON

22

Objective

A multiA multi--angular and multiangular and multi--temporal study of temporal study of vegetation and soils using CHRIS vegetation and soils using CHRIS hyperspectralhyperspectral datadata

1.To examine the variations in spectral directional 1.To examine the variations in spectral directional reflectance of vegetation (cotton)reflectance of vegetation (cotton)

2.To study spatial variability of soil spectra2.To study spatial variability of soil spectra..3.To correlate a selection of indices from different times 3.To correlate a selection of indices from different times

and different angular measurementsand different angular measurements

Goals:

33

Twynam (Colly cotton)

Field Study Site

Centered at 148°51’ E and 29°36’ S

44

Images and Spectral Data Collection

1. CHRIS 1. CHRIS 2. Leaf/Cotton Spectra 2. Leaf/Cotton Spectra 3. Soil Spectra3. Soil Spectra

Spectra for Cotton Varities

010

2030405060708090

100

400 500 600 700 800 900 1000

W ave lenght(nm)

Ref

lect

ance

(%)

55

METHODOLOGY METHODOLOGY Image analysis & classification

Multi-angular & geometric Data

Multiangle Spectra Characteristics

Spectral ID

Multi-angular Multi-temporalHyper spectral dataset Imagery

Reduced Spatial/Spectral Data

Identified & Mapped data

Spectrally Characterized dataClass SeparabilityAnalysis

Ground TruthAccuracy Assessed Data

Statistical Analyses

Results

Vegetation & Soil Spectra

Field Multi-angular Multi-temporal Hyperspectral

Dataset

Calibration

66

� Statistical analysis of image data(surface reflectance)

� Spatial analysisRegistration/geo-correction, ROI

�� Spectral Analysis Spectral Analysis -- EndEnd--member collectionmember collection-- Spectral library creation Spectral library creation -- Classification/spectral Mapping Classification/spectral Mapping ––

e.ge.g MNF, SAM, PPI, Conti. Removed,MNF, SAM, PPI, Conti. Removed,-- Index calculationsIndex calculations

�� MultiMulti--temporal analysistemporal analysis�� Functional Data Analysis (FDA)Functional Data Analysis (FDA)

CHRIS Multi-Angle Image Data Analysis

5 images

2

77

Assumption of using MultiAssumption of using Multi--angular CHRISangular CHRIS

�� Different angle of acquisition has the different Different angle of acquisition has the different response and illumination.response and illumination.

�� MultiMulti--angular image can effect the spectral angular image can effect the spectral analysis results. analysis results.

�� CHRIS can be treated as Functional DataCHRIS can be treated as Functional Data�� How BRDF effect on vegetation and soil.How BRDF effect on vegetation and soil.

?? How this effect, let we explore early findings the How this effect, let we explore early findings the dataset analysis results. dataset analysis results.

88

Radiance

Reflectance

ReflecReflec..+EFFORT+EFFORT

-- Atmospheric correction Atmospheric correction

f

INITIAL RESULTS

99

-- MultiMulti--angular Data Plot (soil)angular Data Plot (soil)

1010

55deg 36deg 0deg -36deg

-55deg

- Multi-angular Spectral Signature

Multi-angular Vegetation Spectra

0.0

0.2

0.4

0.6

0.8

1.0

440 540 640 740 840 940

Wavelength(nm)

Ref

lect

ance

Mean_55

Mean_36

Mean_0

Mean_min36

Mean_min55

Multi-angle soil Spectra

0.0

0.1

0.1

0.2

0.2

0.3

0.3

440 540 640 740 840 940

Wavelength

Ref

lect

ance

Mean_55

Mean_36

Mean_0

Mean_min36

Mean_min55

1111

-- Spectral Spectral EndmemberEndmember Mapping Mapping

�� Minimum Noise Fraction (MNF) to determine inherent image Minimum Noise Fraction (MNF) to determine inherent image datadata

�� Pixel Purity Index (PPI) to determine “pure spectra pixels”Pixel Purity Index (PPI) to determine “pure spectra pixels”�� Spectral Angle Spectral Angle MapperMapper (SAM)(SAM) to classify spectra to match to classify spectra to match

referencereference�� Mixture Tuned Matched Filtering (MTMF) to performs Mixture Tuned Matched Filtering (MTMF) to performs

matched filtering to addmatched filtering to add “infeasibility image result”“infeasibility image result”

1212

+55deg

+36deg

0deg

-36deg

- 55deg

MNF Eigenvalue plot

- Minimum Noise Fraction (MNF) for first 3 bands

3

1313

0

5

10

15

20

25

30

35

40

0 2 4 6 8 10Factor/Band

Tota

l Var

ianc

e (%

)

+55deg

+36deg

0deg

-36deg

-55deg

Percentage of Total Variance of First 10 factors of 60 MNF Bands

- MNF continued…

1414

MultiMulti--angle CHRIS Datasetangle CHRIS Dataset

+55o+36o

0o

-36o

-55oPixel Purity Index

n-D Visualizer

- PPI & n-D Visualizer

1515

- Continuum Removal of Multi-angle dataset

-- Continuum Removal of Continuum Removal of MultiMulti--angle datasetangle dataset

1616

+55, +36, 0, -36, -55

1.0372.685-55deg

1.2832.780-36deg

1.4553.0740deg

1.2402.882+36deg

1.2692.907+55deg

Std. DevMean ValueView Angle

SAM Classification

K-Mean

1717

-- Spectral Variation Within Paddocks (Soil)Spectral Variation Within Paddocks (Soil)

Field Spectra (ASD)CHRIS ImageCHRIS Image

12

3

45

1818

--MultiMulti--temporal Selected Index Valuetemporal Selected Index Value

RSVI

-8

-7

-6

-5

-4

-3

-2

-1

0

24-Jul 14-Jan 6-Jul 27-Dec 18-Jun 9-Dec

Temporal

RV

SI

Val

ue

0deg

36deg

55deg

m36deg

m55deg

Simple Rasio

0

2

4

6

8

10

12

24-Jul 14-Jan 6-Jul 27-Dec 18-Jun 9-Dec

Tempral

SR

Val

ue

0deg

36deg

55deg

m36deg

m55deg

4

1919

NDVI Ratio

y = 0.1654x + 0.4942R2 = 0.8335

0.60

0.61

0.62

0.63

0.64

0.65

0.66

0.67

0.00 0.20 0.40 0.60 0.80 1.00 1.20

Red-Edge Ratio

y = -0.1116x + 4.5486R2 = 0.3254

0

12

34

56

78

0 2 4 6 8

Lab. Red_edge Value

Fie

ld R

ed

-ed

ge

val

ue

RNVI Value

y = -0.4464x + 5.3944R2 = 0.3615

0

1

2

3

4

5

6

-1 0 1 2 3 4

Lab. value

Fie

ld v

alu

e

NDVI Ratio

y = 0.1654x + 0.4942R2 = 0.8335

0.60

0.61

0.62

0.63

0.64

0.65

0.66

0.67

0.00 0.20 0.40 0.60 0.80 1.00 1.20

Correlation of Vegetation Indexes for field spectra,laboratory spectra

2020

CONCLUSIONCONCLUSIONCHRIS MultiCHRIS Multi--angle dataset:angle dataset:

�� The initial investigation has shown the The initial investigation has shown the ‘angular‘angular--signature’ variation and well signature’ variation and well explored.explored.

�� Provide alternative classificationsProvide alternative classifications�� Could provide better classification resultsCould provide better classification results�� Can be used for BRDF studiesCan be used for BRDF studies�� Data function analysis possibleData function analysis possible

2121

The “Hyperspectral Cube” of Colly Cotton Site

HyMap (126 channels)Airborne Hyperspectral

“CHRIS” (62 Channels)Space-borne Hyperspectral