Multispectral Images Tetracam Agriculture Digital to Estimate Nitrogen and Grain Yield of Rice at Different Growth Stages

Abstract

Several methods are available to monitor the nitrogen content of rice during its various growth stages.
However, monitoring still requires a quick, simple, accurate and inexpensive technique that needs to be
developed. In this study, Tetracam Agriculture Digital Camera was used to acquire high spatial and
temporal resolution images to determine the status of N and predict the grain yield of rice (Oryza sativa
L.). Twelve pots of rice were subjected to four different N treatments (0, 125, 175 and 250 kg ha-1). Three
replicates were arranged in a randomized complete block design to determine the status of N and
predict rice yield. The images were captured at different growth stages (i.e., tillering, panicle initiation,
booting and heading stage) of rice in each pot. N and grain yield were significantly correlated with NDVI
(R2 = 0.78) and GNDVI (R2 = 0.88), especially at the panicle initiation and booting stages, respectively.
The study demonstrated the suitability of using the Tetracam images as a sensor for estimating
chlorophyll content and N. Moreover, the findings showed that the images revealed their potential use
in forecasting grain yield at different growth stages of rice.
Key Words: grain yield, nitrogen, rice, Tetracam agriculture digital camera, vegetation indices

Full text

116
INTRODUCTION
Nitrogen (N) is one of the most critical nutrients for crop
production in the world. It is also the nutrient element
applied in the largest quantity for most annual crops
(Huber and Thompson 2007). Fageria et al. (2008)
reported that N was responsible for 85% of the variation
in rice grain yield in Brazilian Inceptisol. In the
developing countries, intensive agricultural production
systems have increased the use of N fertilizer in an effort
to produce and sustain high crop yield, although only 20–
35% of N fertilizer is recovered by the crop because of
losses in several ways (Ponnamperuma and Deturck
1993). It means that N represents the largest part in
fertilizer variable input costs especially in paddy fields
with a double cropping system. On the other hand, lack
of the appropriate amount of N can cause either N stress
or N loss on the crop or environmental pollution, which
is of great concern to the public. Thus, appropriate N
management can help to improve overall rice production
to meet address the need for food security as well as
prevent environmental pollution.
Several different methods are available for assessing
the N status of the crop. One of these methods is tissue
analysis, an example of which is determining the
Kjeldahl N. It is a direct and accurate way of detecting
the N status of a crop, but it is time-consuming and
operators are required. Another technique is use of the
Soil Plant Analysis Development (SPAD) meter to
estimate N status from the measurement of leaf
transmittance in two wavebands centred at 650 nm and
940 nm (e.g., Gholizadeh et al. 2009). Although this
estimation method can be faster and cheaper than
analysis conducted in the laboratory, it still requires field
references, which are derived from laborious procedures
of leaf measurement in the field (Confalonieri et al.
2006).
Multi-Spectral Images Tetracam Agriculture Digital Camera to
Estimate Nitrogen and Grain Yield of Rice at Different Growth Stages
M. M. Saberioon
*, 1
, M. S. M. Amin
1
, W. Aimrun
2
, A. R. Anuar
3
and A. Gholizadeh
1
1
Centre of Precision Farming Technology, Faculty of Engineering, Universiti Putra Malaysia, Serdang, 43400
Selangor, Malaysia
2
Department of Biological and Agricultural Engineering, Faculty of Engineering, Universiti Putra Malaysia
3
Department of Land Management, Faculty of Agriculture, Universiti Putra Malaysia
*
Author for correspondence; e-mail: saberioon@gmail.com; Tel.: + 60 17 6161071; Fax: +60 38 6566061
Several methods are available to monitor the nitrogen content of rice during its various growth stages.
However, monitoring still requires a quick, simple, accurate and inexpensive technique that needs to be
developed. In this study, Tetracam Agriculture Digital Camera was used to acquire high spatial and
temporal resolution images to determine the status of N and predict the grain yield of rice (Oryza sativa
L.). Twelve pots of rice were subjected to four different N treatments (0, 125, 175 and 250 kg ha-1). Three
replicates were arranged in a randomized complete block design to determine the status of N and
predict rice yield. The images were captured at different growth stages (i.e., tillering, panicle initiation,
booting and heading stage) of rice in each pot. N and grain yield were significantly correlated with NDVI
(R2 = 0.78) and GNDVI (R2 = 0.88), especially at the panicle initiation and booting stages, respectively.
The study demonstrated the suitability of using the Tetracam images as a sensor for estimating
chlorophyll content and N. Moreover, the findings showed that the images revealed their potential use
in forecasting grain yield at different growth stages of rice.
Key Words: grain yield, nitrogen, rice, Tetracam agriculture digital camera, vegetation indices
Abbreviations: ADC – agriculture digital camera, DAP – days after planting, GNDVI – Green-Normalized Difference
Vegetation Index, L – Optimal for Adjusting Soil Optical Properties, LNC – leaf nitrogen concentration, NDVI –
Normalized Difference Vegetation Index, NIR – near infrared, RCBD – randomized complete block design, RS –
remote sensing, SAVI – Soil-Adjusted Vegetation Index
ISSN 0031-7454 PHILIPP AGRIC SCIENTIST
Vol. 96 No. 1, 116–121
March 2013
The Philippine Agricultural Scientist Vol. 96 No. 1 (March 2013)

117
Researchers also estimate N status remotely. Remote
Sensing (RS) with aerial images is used to assess N over
the entire fields. Shou et al. (2007) used the high
resolution satellite to evaluate the N status of winter
wheat, which proved that near infrared (NIR) reflectance
has good correlation with sap N concentration, but low
correlation with shoot total N concentration and shoot
biomass. Limitations of these platforms for commercial
use on individual fields are the high cost of images, the
low temporal resolution of satellite images and the short
availability of usable data from satellite images because
of clouds and shadows of clouds. Moreover, the accuracy
of such a technique appears to be easily affected by low
resolution and obvious soil background noises (Broge
and Leblanc 2001). Airborne sensors offer much greater
flexibility than satellite platforms by being able to
operate under clouds and by having a much finer spatial
resolution (Lamb and Brown 2001). However, this type
of imagery is still costly when dedicated ‘mobilization’
of the aircraft is required, especially for remote locations
and repeated data acquisition needs. Therefore, a rapid,
accurate, simple and inexpensive method to monitor a
crop’s N status in the field with high temporal resolution
would be helpful to make rice production more
environmentally and economically sound.
Studies of RS applications in agriculture have used
Vegetation Indices (VIs) to evaluate crop condition. VIs
are easily used and would be an accurate and simple
method in digital image analysis for determining N status
in crops (Hansen and Schjoerring 2003). These indices
are used to enhance the vegetation features and also
reduce the influence of exogenous factors (Daughtry et
al. 2000; Xue et al. 2004).
Tetracam Agriculture Digital Camera (ADC)
(Tetracam, Inc., Chatsworth, Cal.) is one of the ground
platform remote sensors which has been re-engineered
recently to be the simplest and most flexible visible and
near infrared (NIR) camera. It can capture images in
visible light wavelengths longer than 520 nm and near-
infrared wavelengths up to 920 nm. Recently, Swain et
al. (2010) used the Tetracam ADC in paddy fields to
determine the total biomass and to predict the yield as a
function of nutrient status.
Leaf N concentration (LNC) can be determined by
using spectraradiometry from compiled bands of 620 and
760 or 400,620 and 880 nm (Shibayam and Akiyama
1989). Tetracam, which has the potential for determining
the amount of N in crops, covers these wavelengths.
Furthermore, this lightweight camera can be used under
cloudy conditions to monitor the status of N after
planting or during growing stages, with high spatial and
temporal resolution.
This study evaluated the capability of Tetracam ADC
in monitoring N status and estimating yield in paddy
fields on a canopy scale as a fast, inexpensive and simple
technique by using different VIs during the different
growth stages of a crop. Knowledge of N status before
applying fertilizer will enable farmers to practice
precision farming in rice cultivation.
MATERIALS AND METHODS
Theoretical Calculation
In this study, the three most common vegetation indices
were used, namely, Normalized Difference Vegetation
Index (NDVI), Green-Normalized Difference Vegetation
Index (GNDVI) and Soil Adjusted Vegetation Index
(SAVI). NDVI is expressed as follows:
(1)
where NIR is near infrared and R is red.
GNDVI is more highly correlated with final grain
yield in corn (Shanahan et al. 2001). It can be a
promising index for determining the status of N uptake in
wheat during its growth stages (Moges et al. 2005).
GNDVI can be calculated based on the equation
GNDVI (2)
where NIR is near infrared and G is green.
Another index, SAVI, has been developed to
minimize the background influences since NDVI is
affected by soil brightness. SAVI is formulated using the
equation (3)
where NIR is near infrared, R is red and L is the function
of vegetation density.
The value of L is critical in minimizing the influence
of backgrounds in vegetation reflectance. Based on a
suggestion by Huete (1988), L with the value of 0.5 is
optimal for adjusting soil-optical properties.
Field Preparation
The field experiment was conducted at Field No. 2 of the
Universiti Putra Malaysia. The soil in the pots was
obtained from the Tanjung Karang area within Block C
from a major soil series, namely, Telok series (Typic
Sulfaquept). The experiment comprised 12 pots
measuring 3 × 3 m. Rice (Oryza sativa L.) cv. MR-219
was sown in a completely randomized block designed
with three replicates and grown under flooded conditions.
In addition, four different levels of N fertilization were
applied to the rice cultivar, namely, a control with no
fertilizer (N0), and three additional N levels [N1: 22.1,
31.45, 22.95 and 8.5 kg ha
-1
; N2: 44.2, 62.9, 45.9 and 17
Tetracam for Estimation of Nitrogen and Grain Yield M. M. Saberioon et al.
The Philippine Agricultural Scientist Vol. 96 No. 1 (March 2013)

118
kg ha
-1
; and N3: 65, 92.5, 67.5 and 25 kg ha
-1
, which
were applied at the three leaf-stage (15–20 d after
planting, DAP), tillering (35–40 DAP), panicle formation
(55–60 DAP) and grain formation stages (65–70 DAP),
respectively].
Data Acquisition
Vertical images from the central zone of the pots during
tillering, panicle initiation, booting and heading stages
were obtained with a Tetracam ADC (Tetracam Inc.,
Chatsworth, CA, USA). All images were taken between
12:00 and 13:00 on clear days for comparable solar angle
and light intensity, and stored as Raw-8bit image (RAW)
format. The camera had been calibrated by capturing the
images from the Teflon calibration tag under the same
lighting conditions as the images under study.
For a comparison of image acquisition, a SPAD-502
chlorophyll meter (Minolta, Osaka, Japan) was used to
measure the chlorophyll concentration of rice at the
center of the youngest fully developed leaves. Ten
measurements of leaves from each pot were taken and an
average SPAD value for each pot was calculated at each
sampling time as the final value of this rice sample and as
the reference concentration of its N status. Grain yield
data were collected after 110 d. The yield samples were
air-dried and threshed for grains with hull, and the
weights were expressed on the basis of 14% moisture
content.
Image Processing
Images captured by the Tetracam ADC camera were
uploaded to the PixelWrench2 software (Tetracam, Inc,
Chatsworth, Cal.) for deriving the vegetation indices
(VIs). Furthermore, NDVI, Green-NDVI and SAVI were
produced for each image, from which the averages VIs
were estimated.
RESULTS AND DISCUSSION
Estimation of N and Chlorophyll Content Using
Vegetation Indices
Pearson’s correlation between the indexed and SPAD
readings (Table 1) shows that the indices had no
significant correlation with SPAD at the tillering stage,
but all of them showed a significant positive correlation
with the SPAD readings (P = 0.01) at the panicle
initiation and booting stages. Moreover, Table 1 shows
the positive correlation of SPAD readings with only the
NDVI and SAVI at the heading stage. The highest
correlation could be seen between the vegetation indices
and the SPAD readings at panicle initiation. Moreover,
among the indices, SAVI had the highest correlation with
the SPAD readings during the same growth stage (r =
0.896). All three indices generally showed a downward
trend in the correlation with the SPAD readings. In other
words, the correlation between SPAD and NDVI,
GNDVI and SAVI decreased with the growth stage.
The regression model was developed for SPAD with
NDVI, GNDVI and SAVI values in SPSS (IBM
®
SPSS
®
software). The results indicated a good fit (R
2
= 0.78,
RMSE = 2.59) for NDVI at panicle initiation (Table 2).
In general, the indices had the highest correlation with
SPAD during the panicle initiation and booting stages.
Estimation of Grain Yield Using Vegetation Indices
Pearson’s correlation between grain yield and indices
(Table 3) shows a highly significant positive correlation
between indices and grain yield at both panicle initiation
and booting stages (P = 0.01). The highest correlation
could be seen between GNDVI and grain yield at the
booting stage. There was an upward trend in the
correlation between the grain yield and vegetation indices
from panicle initiation to the booting stage. This increase
suggests that the booting stage could be the most
appropriate time to forecast the grain yield by using VIs.
Furthermore, a regression model was developed for both
grain yield and VIs. The results (Table 4) showed a good
regression between GNDVI and grain yield at booting
(R
2
= 0.88, RMSE = 2.33). The estimation models were
developed for this study by using the SPSS in order to
gain good fit models for N estimation. The fit estimation
regression model between SPAD and VIs can be found in
Table 5. Also, the fit estimation regression model
between VIs and grain yield is shown in Table 6.
CONCLUSION
Tetracam ADC was used for capturing multi-spectral
images over rice canopy to estimate the chlorophyll
content, N and grain yield. We found that Tetracam ADC
may be a promising sensor for capturing images to
determine N status and predict yield in rice. The relation
between VIs and SPAD readings also showed the
applicability of the Tetracam ADC images for
determining the status of N in rice. The linear regression
model demonstrated a good fit (R
2
= 0.78) for estimating
N in rice using Tetracam ADC based on NDVI and
GNDVI. The linear model (R
2
= 0.88) suggested that rice
Table 1. Correlation between vegetation index (VI) and
soil plant analysis development (SPAD) reading at
various growth stages of rice.
Growth Stage Days After
Planting NDVI GNDVI SAVI
Tillering 51 0.572
n.s.
0.567
n.s.
0.571
n.s.
Panicle initiation 60 0.887
**
0.886
**
0.896
**
Booting 79 0.785
**
0.783
**
0.749
**
Heading 93 0.789
**
−0.356
n.s.
0.654
**
**
Significant at 0.01.
n.s.
No significant correlation
NDVI – normalized difference vegetation index, GNDVI – green-
normalized difference vegetation index, SAVI – soil adjusted vegetation
index
The Philippine Agricultural Scientist Vol. 96 No. 1 (March 2013)
Tetracam for Estimation of Nitrogen and Grain Yield M. M. Saberioon et al.

119
yield could be predicted by using GNDVI derived from
images obtained by the Tetracam ADC. Our study also
confirmed that the best growth stages for measuring the
status of N in rice would be the panicle initiation and
booting stages. Further study is recommended to evaluate
the capability of other nutrients, which have important
roles during the different growth stages of rice in paddy
fields.
REFERENCES CITED
BROGE NH, LEBLANC E. 2001. Comparing prediction power
and stability of broadband and hyperspectral vegetation
indices for estimation of green leaf area index and canopy
chlorophyll density. Remote Sens Environ 76:156–172.
CONFALONIERI R, STROPPIANA D, BOSCHETTI M,
GUSBERTI D, BOCCHI S, ACUTIS M. 2006. Analysis of
rice sample size variability due to development stage,
nitrogen fertilization, sowing technique and variety using
the visual jackknife. Field Crop Res 97:135–141.
DAUGHTRY CST, WAITHALL CL, KIM MS, DE
COLSTOUN EB, McMURTREY III JE. 2000. Estimating
corn leaf chlorophyll concentration from leaf and canopy
reflectance. Remote Sens Environ 74: 229–239.
FAGERIA NK, BALIGAR VC, LI YC. 2008 the role of
nutrient efficient plants in improving crop yields in the
twenty first century. J Plant Nutr 31(6):1121–1157.
GHOLIZADEH A, AMIN MSM, ANUAR AR, AIMRUN W.
2009. Evaluation of leaf total nitrogen content for nitrogen
management in a Malaysian paddy field by using soil plant
analysis development chlorophyll meter. Am J Agric Biol
Sci 4 (4): 278–282.
HANSEN PM, SCHJOERRING JK. 2003. Reflectance
measurement of canopy biomass and nitrogen status in
wheat crops using normalized difference vegetation indices
and partial least squares regression. Remote Sens Environ
86:542–553.
Table 2. Regression between vegetation index (VI) and
soil plant analysis development (SPAD) reading at
different growth stages of rice.
Growth
Stage
NDVI GNDVI SAVI
R2
Root
Mean
Square
Error
R2
Root
Mean
Square
Error
R2
Root
Mean
Square
Error
Panicle
initiation 0.78 2.59 0.78 2.60 0.70 3.04
Booting 0.71 2.18 0.75 2.05 0.77 1.94
Heading 0.62 2.03 - - 0.42 2.50
NDVI – normalized difference vegetation index, GNDVI – green-
normalized difference vegetation index, SAVI – soil adjusted vegetation
index, R
2
– regression
Table 3. Correlation between vegetation index (VI) and
grain yield at different growth stages of rice.
Growth
Stage
Days
After
Planting NDVI GNDVI SAVI
Tillering 51 0.581
*
0.577
*
0.580
*
Panicle
initiation 60 0.859
**
0.857
**
0.803
**
Booting 79 0.915
**
0.940
**
0.939
**
Heading 93 0.642
*
−0.485
n.s.
0.778
**
**
Significant at 0.01.
*
Significant at 0.05.
n.s.
No significant correlation
NDVI – normalized difference vegetation index, GNDVI –
green-normalized difference vegetation index, SAVI – soil
adjusted vegetation index
Table 4. Regression between vegetation index (VI) and
grain yield at various growth stages of rice.
Growth
Stage
NDVI GNDVI SAVI
R
2
Root
Mean
Square
Error
R
2
Root
Mean
Square
Error
R
2
Root
Mean
Square
Error
Panicle
initiation 0.73 3.50 0.73 3.52 0.64 4.07
Booting 0.83 2.76 0.88 2.33 0.88 2.39
Heading 0.41 5.2
NDVI – normalized difference vegetation index, GNDVI – green-
normalized difference vegetation index, SAVI – soil adjusted vegetation
index, R
2
– regression
Table 5. The curve fit estimation regression model between soil plant analysis development (SPAD) reading and
vegetation indices.
Growth
Stage NDVI GNDVI SAVI
Panicle initiation SPAD= 87.36*NDVI-30.19 SPAD= 88.595*GNDVI-30.828 SPAD= 84.615*SAVI-28.20
Booting SPAD= 40.412*NDVI+29.76 SPAD= 46.323*GNDVI+29.33 SPAD= 41.956*SAVI+29.40
Heading SPAD = 52.07*NDVI+26.75 SPAD= 30.918*SAVI+31.03
SPAD – Soil Plant Analysis Development, NDVI – Normalized difference vegetation index, GNDVI – Green normalized difference vegetation index, SAVI –
Soil adjusted vegetation index
Table 6. The curve fit estimation regression model between grain yield and vegetation indices
Growth
Stage NDVI GNDVI SAVI
Panicle initiation Yield= 103.093*NDVI-63.013 Yield= 104.447*GNDVI-63.691 Yield= 98.58*SAVI-59.706
Booting Yield= 72.258*NDVI+4.171 Yield= 83.23*GNDVI+3.360 Yield= 74.086*SAVI+3.692
Heading Yield= 62.694*SAVI+0.213
SPAD – Soil Plant Analysis Development, NDVI – Normalized difference vegetation index, GNDVI – Green normalized difference vegetation index, SAVI –
Soil adjusted vegetation index
The Philippine Agricultural Scientist Vol. 96 No. 1 (March 2013)
Tetracam for Estimation of Nitrogen and Grain Yield M. M. Saberioon et al.

120
HUBER DM, THOMPSON IA. 2007. Nitrogen and plant
disease. In: Datnoff LE, Elmer WH, Huber DM, editors.
Mineral Nutrient and Plant Disease. St. Paul, MN, USA:
The American Phytopathological Society. p. 31–44.
HUETE AR. 1988. A soil vegetation adjusted index (SAVI).
Remote Sens Environ 25 (1988): 295–309.
LAMB DW, BROWN RB. 2001. Remote sensing and mapping
of weeds in crops. J Agric Eng Res 78(2): 117–125.
MOGES SM, RAUN WR, MULLEN RW, FREEMAN KW,
JOHNSON GV, SOLIE JB. 2005. Evaluation of green, red,
and near infrared bands for predicting winter wheat
biomass, nitrogen uptake, and final grain yield. J Plant
Nutr 27(8): 1431–1441.
PONNAMPERUMA FN, DETURCK P. 1993. A review of
fertilization in rice production. Int Rice Comm Newslett
42:1–12.
SHANAHAN JF, SCHEPERS JS, FRANCIS DD, VARVEL
GE, WILHELM WW, TRINGE JS, SCHLEMMER MR,
MAJOR DJ. 2001. Use of remote sensing imagery to
estimate corn grain yield. Agron J 93: 583–589.
SHIBAYAM M, AKIYAMA T. 1989. Estimating grassland
phytomass production with near-infrared and mid-infrared
spectral variables. Remote Sens Environ 30(3): 257–261.
SHOU L, JIA L, CUI Z, CHEN X, ZHANG F. 2007. Using
high-resolution satellite imaging to evaluate nitrogen status
of winter wheat. J Plant Nutr 30(10): 1669–1680.
SWAIN KC, THOMSON HP, JAYASURIYA HPW. 2010.
Adoption of an unmanned helicopter for low altitude
remote sensing to estimate yield and total biomass.
ASABE J 53(1): 21–27.
XUE L, CAO W, LUO W, DAI T, ZHU Y. 2004. Monitoring
leaf nitrogen status in rice with canopy spectral reflectance.
Agron J 96: 135–142.
The Philippine Agricultural Scientist Vol. 96 No. 1 (March 2013)
Tetracam for Estimation of Nitrogen and Grain Yield M. M. Saberioon et al.