Skip to main content
ARS Home » Plains Area » Manhattan, Kansas » Center for Grain and Animal Health Research » Stored Product Insect and Engineering Research » Research » Publications at this Location » Publication #399059
Research Project: Advancing Technologies for Grain Trait Measurement and Storage Preservation

Location: Stored Product Insect and Engineering Research

Title: Prediction of sorghum oil and kernel weight using near-infrared hyperspectral imaging

Author
item MENDOZA, PRINCESS TIFFAN - Kansas State University
item Armstrong, Paul
item PEIRIS, KAMARANGA - US Department Of Agriculture (USDA)
item SILIVERU, KALIRAMESH - Kansas State University
item BEAN, SCOTT - US Department Of Agriculture (USDA)
item Pordesimo, Lester

Submitted to: Cereal Chemistry
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 2/22/2023
Publication Date: 2/24/2023
Citation: Mendoza, P.D., Armstrong, P.R., Peiris, K.H., Siliveru, K., Bean, S.R., Pordesimo, L.O. 2023. Prediction of sorghum oil and kernel weight using near-infrared hyperspectral imaging. Cereal Chemistry. 100(3):775-783. https://doi.org/10.1002/cche.10656.
DOI: https://doi.org/10.1002/cche.10656

Interpretive Summary: The growing demand for sorghum grain has increased interest for breeders, producers, and processors, to focus as on grain quality beyond grain yield, and include composition such as oil content and kernel size or weight. Near-infrared spectroscopy has been used as a reliable tool for fast and non-destructive evaluation of major quality properties of grains as is a substitute time-consuming and costly wet chemistry methods. Near-infrared hyperspectral imaging (NIR HSI), which combines spectroscopy and machine vision, is a more advanced method that is now being used in the food industry to predict quality parameters at low concentrations and can reveal the wide variability within the sample. In this study, NIR hyperspectral imaging was explored as a method to predict single-seed weight and oil content of sorghum grains. Comparisons were made between hyperspectral imaging and other techniques for measuring grain weight and oil content. Model evaluation showed satisfactory results that were comparable to those from single-kernel near-infrared reflectance instrument using the same set of samples. This study showed the potential of hyperspectral imaging as a quality control method for sorghum grains and as a tool to investigate variance in oil content in different sorghum lines which would be beneficial to sorghum breeders.

Technical Abstract: This study was performed to explore near-infrared hyperspectral imaging (NIR HSI) as a non-destructive and rapid method to predict single seed weight and oil content of sorghum grains. Sorghum grains (76 samples) from 2019 to 2020 were scanned with a Resonon Pika NIR-640, a line scan push-broom reflectance hyperspectral camera, in the 900 to 1700 nm range. Calibration models were developed using partial least squares (PLS) regression. The final model for oil from NIR HSI spectra achieved a 0.19% standard error of calibration (SEC) and 0.21% standard error of prediction (SEP) at 10 PLS factors, while kernel weight prediction obtained 40 mg SEC and 50 mg SEP at 11 PLS factors. The results from the NIR HSI instrument were comparable to those from single-kernel near-infrared reflectance instrument (SKNIR) using the same set of samples. This study showed the potential of hyperspectral imaging as a quality control method for sorghum grains, specifically for oil content, which could be beneficial for sorghum breeders, growers and processors.