DEVELOPMENT OF A PROTEIN SENSOR FOR COMBINE HARVESTERS
L. THYLÉN, P.A. ALGERBO
JTI, Uppsala, Sweden
E-mail : lars.thylen@jti.slu.se, per-anders.algerbo@jti.slu.se
ABSTRACT
Grain quality within fields shows a large variability, and by sorting the grain the overall
economic value can be increased. The objectives of this study were to develop an on-line
protein sensor for combine harvesters, and to test the system in the field.
A modified NIT-instrument, measuring 14 wavelengths (893 – 1045 nm) of transmitted light
as well as of grain and sensor temperature, was installed within a sampling unit attached to the
clean grain elevator of a combine harvester. The sensor was calibrated for protein and
moisture content. For wheat the correlations (R²) between lab and sensor readings were 0.99
for moisture and 0.93 for protein. When using the system in the field, about 30 samples per
hectare were measured. The collected data had a good spatial dependency and the average
kriging standard deviation within the fields was 0.12%.
INTRODUCTION
Several studies have shown that crop quality within fields is highly variable (Mulla et al.,
1992; Dawson, 1996; Reyns et al., 1999). This variability can have a major effect on the
economic value of the crop. However, by sorting the grain the overall economic value can be
increased (Thylén et.al., 1999; Stafford, 1999). To be able to sort the grain into different
fractions, a sensor detecting the grain quality either before or during the harvest is needed.
Traditionally, samples are collected during the harvest and later analysed in the lab. This is
usually done by Kjeldahl anlysis or with a NIR/NIT instrument. The possibility to use NIR to
measure protein content in whole grain was investigated by Williams et al (1985). An
approach for measuring grain protein content on-line by the use of near infrared reflectance
was described by Engel et al. (1997).
The aims of this study were to develop a system that can measure protein content on-line in a
combine harvester, and to test the system during harvest.
MATERIAL AND METHODS
NIT-sensor
In this project we used a grain protein sensor which operates in the short-wavelength NIR,
from 893 to 1045nm. The portable instrument, Zeltex ZX-50 (Rosenthal et al, 1999), can be
calibrated to analyse moisture, protein and other constituent contents from the absorption
spectra of the grains being tested. Besides measuring the spectra, the ambient and sample
temperatures are measured. These data are used to calculate moisture and protein content of
1
the sample measured. When using the instrument in the lab, a sample cup is filled with grain
and inserted into the sensor. The sample is illuminated and the transmitted light detected.
Sampling and control device
The protein sensor was installed in a sampling device mounted on the side of the clean grain
elevator. To avoid problems caused by vibrations, the sensor was built into a unit, which
dampened the vibrations. Vibrations do not affect the sensor directly, but moving kernels
during measurements will cause bad data. Cutting a hole in the bottom of the NIT sensor and
letting grain pass through the sample cup made it possible to intermittent measure grain
protein content. Two flaps controlled the grain flow through the sensor. A level sensor
between the flaps made sure that the sample cup was properly filled during light exposure
(figure 1). During harvesting there is a risk that the sample cup will be scratched, therefore
affecting the amount of light transmitted. To avoid this problem, every fifth reading was
carried out with an empty sample cup thus taking the wear into account.
FIGURE 1. The sampling device was mounted on the side of the clean grain elevator. Two
flaps controlled the flow of grain through the NIT- sensor. The level sensor made sure that the
sample cup was properly filled during light exposure.
A control unit located next to the NIT-sensor controlled both the flaps and the NIT-sensor. To
do so the software in the NIT-sensor was modified. The control unit also recorded the speed
of the combine and the header position. When using the system in field all data was
transmitted to a handheld PC in the cab of the combine. The PC was also connected to a GPSreceiver. During the harvest of 2000, two systems were installed on combine harvesters and
tested in the field.
Test bench trials
2
Since the NIT-sensor was modified, the sensor was calibrated with grain samples of known
protein and moisture content. This calibration was performed with the complete samplingdevice. However, the sensor was filled with grain through a small hopper and not by the clean
grain elevator.
RESULTS
Test bench trials
16
Predicted moisture content,
(%)
Predicted protein content, (%)
The calibration of the sensor was performed with wheat and barley and the measured
constituents were protein and moisture content. Examples of the calibration curves for wheat
are shown in figure 2.
15
14
13
12
11
10
9
8
8
10
12
14
16
19
18
17
16
15
14
13
12
11
10
10
Lab protein content, (%)
12
14
16
18
20
Lab moisture content, (%)
FIGURE 2. Results from calibration of one protein sensor with wheat. The graph to the left
shows the protein content, and the graph to the right shows moisture content.
Field trials
The two sensor systems were tested during harvesting in 2000. During normal operation the
system managed about 30 readings per ha. About two percent of the collected data were
erroneous, but these were possible to detect when studying the transmitted data. The collected
data was interpolated, using blockkriging (20*20 m) and local semivariograms, with the
software Vesper (Minasny et al, 1999) and the kriging standard deviation was also calculated.
An example is shown in figure 3. In this case the protein content ranged between 11 and 14.6
% and the average kriging standard deviation was 0.12 %. The moisture content of the grain
was mainly depending of time of day, one example is shown in figure 4.
3
6640000
Protein
%
Yield
ton/ha
Y-coordinate, metres
6639900
6639800
7
6639700
6
6639600
5
6639500
4
6639400
3
14
13
12
11
6639300
1599600
1599800
1600000
1599600
X-coordinate, metres
1599800
1600000
X-coordinate, metres
FIGURE 3. Example of maps showing yield and protein content from Hallby field. The field
(ca 23 ha) was cropped with winter wheat, variety Kosack.
Moisture content, (%)
22
20
18
16
14
12
10
09:00
11:00
13:00
15:00
17:00
19:00
Time
FIGURE 4. The moisture content of the harvested grain decreased during the day, but a slight
increase in moisture content can be seen during the last hour of operation.
During harvesting, two problems occurred with the protein sensor. When combining grain
with a moisture content close to 30%, the sensor got blocked with wet grain. We also had a
mechanical problem with the flap controlling the inlet flow of grain to the sensor.
DISCUSSION
The possibility for on-line measurement of protein content during harvesting has been
discussed for several years. Knowing grain protein quality can be used to sort grain into
different fractions, thereby increasing the economic value of the harvested crop. The
4
information can, in conjunction with yield data, also be used to calculate the nitrogen uptake
with the crop.
In this study we adapted an existing portable grain protein sensor for use on combine
harvesters. The system worked properly during harvest, but the sensor tended to get blocked
when the moisture content was close to 30%.
ACKNOWLEDGEMENTS
This research is funded by The Swedish Farmers’ Foundation for Agricultural Research
(SLF). The authors are thankful for all support from Zeltex.
REFERENCES
Dawson, C.J. (1996) Implications of precision farming for fertilizer application policies.
proceedings No. 391, The Fertilizer Society, 44pp.
Engel, R., Long, D., Carlson, G. (1997) On-the-go grain protein sensing is near. Better crops
with plant food, 81, no 4, pp 20-23.
Minasny, B., McBratney, A.B., Whelan B.M. (1999) VESPER version 1.0. Australian Centre
for Precision Agriculture, McMillan Building A05, The University of Sydney, NSW
2006. (http://www.usyd.edu.au/su/agric/acpa).
Mulla, D.J., Bhatti, A.U., Hammond, M.W., Benson, J.A. (1992) A comparison of winter
wheat yield and quality under uniform versus spatially variable fertilizer management.
Agriculture, Ecosystems and Environment, 38, pp 301-311.
Reyns, P., De Baerdemaeker, J., Ramon, H. (1999) Site-specific relationship between grain
quality and yield, Precision agriculture ’99, Stafford, J.V. (Ed), UK, Sheffield
Academic Press, pp 665-676.
Rosenthal, T., Woycik, H., Kristof,K. (1999) Use of portable Near-infrared grain analyzers in
developing countries, presented at the Ninth international Conferenceon NIR
Spectroscopy, Verona Italy, June 1999.
Stafford, J.V. (1999) An investigation into the within-field spatial variability of grain quality,
Precision agriculture ’99, Stafford, J.V. (Ed), UK, Sheffield Academic Press, pp 353361.
Thylén, L., Algerbo, P.A., Pettersson, C.G. (1999) Grain quality variations within fields of
malting barley, Precision agriculture ’99, Stafford, J.V. (Ed), UK, Sheffield Academic
Press, pp 287-296.
Williams, P.C., Norris, K.H., Sobering, D.C. (1985) Determination of protein and moisture in
wheat and barley by near-infrared transmission. J.Agric-Food-Chem. Washington, D.C.
American Chemical Society. v. 33 (2) pp. 239-244.
5