Chlorophyll Estimation
Using Multi-spectral
Reflectance and Height
Sensing
2004 ASAE/CSAE Annual
International Meeting
Paper Number: 043081
C. L. Jones
N. O. Maness
M. L. Stone
R. Jayasekara
Research Engineer
Professor
Regents’ Professor
Research Engineer
Objectives
• Select “best” wavelength or indices for estimating
chlorophyll content and concentration in spinach
• Non-destructively quantify chlorophyll content and
concentration by combining biomass estimates with
multi-spectral imaging system reflectance data.
• Assess the feasibility of reflectance-based sensing
in estimating chlorophyll concentration in spinach
Paper Number: 043081
Let’s Talk Chlorophyll…
(Content vs. Concentration)
• Content: related to nitrogen
concentration in green vegetation
• Chlorophyll mass per unit ground
surface area or per plant (mg/ha or
mg/plant)
• Indicator of photosynthetic capacity:
productivity of the plant canopy
Paper Number: 043081
Let’s Talk Chlorophyll…
(Content vs. Concentration)
• Concentration: estimate of pigment
concentration – chlorophyll a and b
• Chlorophyll mass per unit mass of plant
material (mg/kg)
• Indicator of plant physiological status:
level of stress
Paper Number: 043081
Previous Work
No single spectral approach offers an
adequate method of estimating
pigment (chlorophyll) concentration in
all plants.
Paper Number: 043081
Experimental Plan
• Consider spinach plants at the single plant
level
• Use chlorophyll content estimates to
determine chlorophyll concentration
(implies a need for biomass estimate)
• Ccontent (mg) = Cconcentration (mg/kg) x biomass
(kg)
• Use ultrasound and digital imagery for
biomass estimate
Paper Number: 043081
Experimental Plan
Estimators used for chlorophyll content
investigation:
– Rgreen : Reflectance at 550 nm ±10 nm
– Rred : Reflectance at 670 nm ±10 nm
– RNIR : Reflectance at 780 nm ±10 nm
– NIR/RED : Ratio of reflectance at 780 and 670 nm ±10 nm
– NIR/GREEN : Ratio of reflectance at 780 and 550 ±10 nm
– NDVI670 = (R780 – R670)/(R780 + R670)
– NDVI550 = (R780 – R550)/(R780 + R550)
Paper Number: 043081
Plant Samples
Two flats- no fertilizer
Two flats adequate fertilizer
Two flats 2x adequate fertilizer
amount
Provided varying biomass and
chlorophyll levels
Sampled at six weeks post
plant
Paper Number: 043081
Data Gathering Overview
• Plants removed from flats
• Individually placed on turntable (0.91
m diameter) 1.04 m below ultrasonic
distance sensor (UDS) for height
estimation
• Turntable rotating at 1.06 rpm
Paper Number: 043081
Data Gathering Overview
• Plants placed under multispectral camera
mounted on tripod for reflectance data and
top view surface area estimate
• Vegetative portion harvested, weighed, and
packed on ice for transport to lab
• Chlorophyll analysis per Inskeep and Bloom
(1985)
Paper Number: 043081
Ultrasonic Distance Sensor
(UDS)
Senix™ Ultra-SPA.
(Senix, Bristol, Virginia)
• 50 kHz, piezoelectric
• 12 degree angle view, signal > 3db
• Accuracy: better than 1% of target
distance
• Laptop computer
• SoftSpan© software
Paper Number: 043081
UDS Response
Spinach Height Profile - Ultrasound
18
Height,cm
16
14
12
10
8
6
4
2
0
1.37 2.74 4.11 5.49 6.86 8.23
9.6
11
12.3 13.7 15.1 16.5 17.8 19.2
Range, cm
Paper Number: 043081
Multispectral Camera
•
•
•
•
•
•
DuncanTech MS3100
550, 670, 780 nm ±10 nm FWHM
Resolution: 1392 x 1040 x 3 sensors
Pixel size: 4.65 x 4.65 microns
14mm focal length lens
RS232 interface with National Instrument’s
PCI-1424 frame grabber
• Dolch ruggedized laptop
Paper Number: 043081
Multispectral Camera
• Mounted on tripod 1 m above target
• Target illuminated with incandescent lighting
• Labsphere Reflectance Calibration Standard
targets: 10%, 50%, 75%, and 99% (Weckler, et
al. 2002)
Paper Number: 043081
Multispectral Camera
• Images processed with Matlab©
• NIR, green, and red reflectance pixel values
identified
ρNIR - ρGreen or Red
NDVI =
ρNIR + ρGreen or Red
• > 0 : plant material, < 0 : background, nonplant
material, NDVI averaged
• Image binarized
• “bwarea” command to calculate vegetation area
Paper Number: 043081
Biomass Estimation
WEST = AMS x HUDS
Where: WEST = estimated plant biomass
AMS = surface area estimate from camera (pixels)
HUDS = estimated plant height from UDS (cm)
PWC = 0.86
Paper Number: 043081
Biomass Estimation
W = 3E-05WEST + 2.0958
r 2 = 0.884
Actual Biomass (W), g
35
30
25
20
15
10
5
0
0.00E+00
2.00E+05
4.00E+05
6.00E+05
8.00E+05
1.00E+06
1.20E+06
Estimated Biomass (WEST), H UDS * AMS
PWC = 0.86
Paper Number: 043081
Chlorophyll Content
Estimation, r2
Chl. Content, mg/plant
300
C
= 1364.5NDVI - 370.41
0.75
Ccontentr2 =vs.
CONT
Estimator
estimator
0.21
250
200
Rgreen
150
Rred
0.10
100
RNIR
0.58
NIR/RED
0.15
NIR/GREEN
0.05
50
0
NDVI670
0
NDVI550
0.1
0.2
0.3
0.75
0.4
0.5
NDVI
0.40
Paper Number: 043081
Chlorophyll Content
Estimation, r2
Estimator
Ccontent vs.
Ccontent vs.
=
17.478(NDVI)(W
)
+
22.82
CCONT(Estimate)
estimatorr2 = 0.910 ESTestimator * WEST
0.21
0.87
350
Rred
Chl. Content, mg/plant
Rgreen
300
0.10
0.86
0.58
0.90
NIR/RED
100
0.15
0.90
50
NIR/GREEN
0.05
0.89
RNIR
250
200
150
NDVI670
NDVI550
0
0.00
2.00
4.00
0.75 8.00
6.00
NDVI*WEST
0.40
10.00
12.00
0.91
14.00
16.00
0.89
Paper Number: 043081
Chlorophyll Concentration
Estimation
•Estimated chlorophyll content
calculated from regression equations
for each band or ratio
•CCONC = CCONT / WEST
•Compared to actual Chlorophyll
concentration
Paper Number: 043081
Chlorophyll Concentration
Estimation, r2
Estimator
Ccontent vs.
estimator
Ccontent vs.
estimator *
WEST
Actual Cconc
vs. Cconc
estimate
Rgreen
0.21
0.87
0.22
Rred
0.10
0.86
0.30
RNIR
0.58
0.90
0.05
NIR/RED
0.15
0.90
0.04
NIR/GREEN
0.05
0.89
0.03
NDVI670
0.75
0.91
0.08
NDVI550
0.40
0.89
0.000004
Paper Number: 043081
Conclusions
• Of the indices tested in this study, NDVI670
multiplied by estimated biomass appears to
provide best estimate of chlorophyll content in
spinach
• Multiplying reflectance ratios and indices by
estimated biomass improved chlorophyll content
estimations
• Reflectance-based remote sensing may not be
the best method for estimating plant pigment
concentrations (chl. a and b): RRED best, r2 = 0.30
Paper Number: 043081
QUESTIONS?
Acknowledgments
Support through funding from the USDA Special
Research Grant Number 2003-06134
Paper Number: 043081