Adapting a relatively low-cost reflectance spectrometer for on-combine sensing of grain protein concentration

https://doi.org/10.1016/j.compag.2020.105467Get rights and content

Highlights

  • A laboratory spectrometer was adapted for use on a combine harvester for mapping grain protein.

  • Predicted grain protein concentration compared favorably with that of reference grain samples.

  • A protein map derived from the low-cost instrument compared with one made from a more expensive instrument.

Abstract

An on-combine, near infrared (NIR) spectrometer is now commercially available for mapping grain protein concentration (GPC) within farm fields. Costing > US$20,000, this specialized instrument may not be affordable for all producers. The objective of this study was to adapt and evaluate a moderately priced reflectance spectrometer (<US$5500, Avantes AvaSpec ULS2048 × 16-USB2) for on-combine sensing of GPC. A “leave-one-out” cross-validation partial least square (PLS) regression model for GPC was developed for flowing grain with the spectrometer’s probe-head attached to the lower end of an inclined trough. Good correlation (r2 = 0.92) was observed between estimated and reference GPC in grain samples when NIR spectra (850–1040 nm) and GPC reference values were used for developing the PLS model. The NIR estimation achieved a standard error of < 5 g of protein kg−1 of grain. An inexpensive external temperature control system was built that stabilized the temperature of the instrument’s input/output converter board within 60 m of activation. Placed on a test stand machine with standard 12.5 cm diameter auger, the instrument predicted GPC to < 5 and < 10 g kg−1 in two and three of four wheat cultivars thus indicating that the model produced on the trough may be applied into predicting GPC on a combine. When placed on a combine, the calibrated instrument produced a protein map that correlated well (r = 0.88) with a map derived from an expensive reflectance spectrometer (US$32,000, Polytec 1721). Thus, a relatively inexpensive NIR reflectance spectrometer may be adapted for mapping GPC across fields.

Introduction

Near infrared (NIR) spectroscopy is a standard technique used in the wheat industry for nondestructive measurement of the protein concentration of whole grains. Multivariate calibration models are utilized to predict grain protein concentration (GPC) once a relationship is established between GPC and diffuse NIR absorption spectra. Caporaso et al. (2018) cite Williams et al., 1978, Norris, 1982 as the first applications of NIR spectroscopy on ground wheat and flour followed by Williams and Sobering (1993) on whole kernels and Delwiche (1998) on single kernels. Today, bench spectrometers for laboratory determination of GPC provide a rapid, reliable means of estimating whole grain quality and assessing dollar value in elevators and other delivery points. Further advances in engineering and manufacturing have made it possible to design ruggedized instruments capable of continuously sensing the grain stream on a combine harvester (Maertens et al., 2004, Taylor and Whelan, 2007).

When linked to a global navigation satellite system (GNSS), thousands of GPC measurements acquired from an on-combine spectrometer can be mapped across a farm field (Long et al. 2008). This information has practical use in precision agriculture. For instance, protein maps have been found to be useful in assessing nitrogen nutrition adequacy for grain yield once a critical protein level is known for a specific wheat class and region (Engel et al., 1999, Goos, 1984, Selles and Zentner, 2001). Combined with site-specific data from a grain yield monitor, mapped values of grain protein can be used to compute the amount of nitrogen needed to replace that removed in the grain at harvest and adjust protein in the next wheat crop to a certain target level (Engel et al., 1999). The rationale is that crops are excellent indicators of soil conditions in the root zone and grain protein is highly correlated with nitrogen fertility.

Several on-combine instruments appeared between 2000 and 2010 (von Rosenberg et al., 2000, Long and Rosenthal, 2005), but were not profitable due to low sales and are no longer available. At that time, potential customers questioned the affordability and usefulness of the technology. Currently, an improved on-combine instrument is available from Next Instruments (Bankstown, NSW, Australia) for around US$23,000. Interest in this instrument is increasing among wheat growers in Australia who wish to segregate grain from areas of high- or low-protein within fields (Leonard, 2017). By blending grain of high or low quality to achieve grain of a higher grade, a greater volume of grain can be sold at a higher price versus leaving the grain bulked and selling it by load (Stewart et al. 2002).

Moderately priced spectrometers (<US$10,000) have come on the market in recent years for laboratory use that offer some of the same features as earlier instruments that were sold for on-combine use (e.g., compactness, solid-state photon detectors, thermal stability, sealed housing). Specialized instruments that are designed for use on combine harvesters may be too costly for many dryland wheat growers in the US who typically gross <$80 ha−1 under semi-arid conditions. The objective of this study was to adapt and evaluate a moderately priced reflectance spectrometer for on-combine sensing of GPC. Affordable instruments would help promote the adoption of this technology for use in precision nitrogen management and grain segregation/blending.

Section snippets

Instrument description

The instrument used in this study was the AvaSpec ULS2048 × 16-USB2 Spectrometer (Avantes, Apeldoorn, The Netherlands) having rugged and field deployable characteristics (Fig. 1). This spectrometer utilizes a Charged Coupled Device (CCD) linear array of 2048 elements (Hamamatsu S11071-1104) to measure diffuse reflectance spectra at 0.7 nm intervals over a wavelength range of 650 to 1150 nm when configured with grating of 600 lines mm−1 and slit of 25 μm. Sensitivity is improved beyond 1000 nm

Grain spectra

Fig. 6A shows raw reflectance spectra of flowing grain samples of “Alpowa” soft white spring wheat as measured by the Avantes instrument on the trough. Spectra are colored by different levels of GPC. No clear absorption band for GPC can be seen. The dominant feature in these spectra is the baseline shift caused by light scattering from random grain orientation. However, the second derivative and binning of these spectra effectively eliminated the baseline drift and revealed strong bands, or

Discussion

On the combine, the insulated container maintained an input/output circuit board temperature of around 23 °C during operation indicating that the insulating effect of air between the spectrometer and the cold plate was adequate even though the energy of the halogen lamp exceeded the demand of the instrument when collecting spectra. According to recorded measurements of the circuit board, the detector temperature could be controlled by enclosing the instrument within an insulated cardboard box

Funding

This work was supported in part by the Oregon Wheat Commission.

CRediT authorship contribution statement

Dan S. Long: Conceptualization, Resources, Supervision, Project administration, Writing - original draft. John D. McCallum: Methodology, Software, Validation, Formal analysis, Investigation, Data curation, Visualization, Writing - original draft.

Declaration of Competing Interest

The authors declared that there is no conflict of interest.

Acknowledgements

Grateful appreciation to anonymous reviewers for helpful comments, Stephen Machado of Oregon State University for providing the spectrometer used in this study, Steve Umbarger for technical support in the field, and Damon Lenski of Avantes for providing technical information on the spectrometer.

References (27)

  • J. Barroso et al.

    Optical sensing of weed infestations at harvest

    Sensors

    (2017)
  • N.B. Büchmann et al.

    Performance of European artificial neural network (ANN) calibrations for moisture and protein in cereals using the Danish near-infrared transmission (NIT) network

    Cereal Chem.

    (2001)
  • N. Caporaso et al.

    Near-infrared spectroscopy and hyperspectral imaging for non-destructive quality assessment of cereal grains

    Appl. Spectrosc. Rev.

    (2018)
  • S.L. Delwiche

    Protein content of single kernels of wheat by near infrared reflectance spectroscopy

    J. Cereal Sci.

    (1998)
  • J.C. Dunlap et al.

    Interpreting activation energies in digital image sensors

    IEEE Trans. Electron.

    (2016)
  • R.E. Engel et al.

    Method for precision nitrogen management in spring wheat: II. Implementation

    Precis. Agric.

    (1999)
  • R.J. Goos

    Post-harvest evaluation of nitrogen management- a new approach for “selling” soil testing to wheat farmers

    J. Agron. Educ.

    (1984)
  • Leonard, E. 2017. On-the-go combine protein sensors increase wheat margins. Future Farming. 18 September, Available at...
  • Long. D.S., and T. Rosenthal. 2005. Evaluation of an on-combine wheat protein analyzer on Montana hard red spring...
  • Long, D.S., R.E, Engel, and M.C. Siemens. 2008. Measuring grain protein concentration with in-line near infrared...
  • D.S. Long et al.

    Nitrogen requirement to change protein concentration of spring wheat in semiarid Pacific Northwest

    Agron. J.

    (2017)
  • K. Maertens et al.

    On-line measurement of grain quality with NIR technology

    Trans. ASAE

    (2004)
  • K. Norris

    Use of near-infrared reflectance spectroscopy and dye-binding techniques for estimating protein in oat groats

    Cereal Chem.

    (1982)
  • Cited by (0)

    View full text