Prediction of the chemical composition of winter grain and maize with near infrared reflectance spectroscopy

Abstract

Near infrared reflectance spectroscopy (NIRS) was evaluated as a tool for the fast and inexpensive prediction of nutritional values of feedstuffs. NIRS calibrations were developed for winter grain ,samples collected over three years in the Western Cape region of South Africa. Winter grains used in the study include oats, barley, triticale and wheat. Calibrations were also developed for maize samples collected throughout South Africa. Winter grain samples were analysed for ash, dry matter (OM), crude protein content (CP), fat content, acid detergent fibre (ADF), neutral detergent fibre (NOF), in vitro organic matter digestibility (IVOMO), lysine and methionine. These results suggested that NIRS can be developed as a rapid and accurate tool for the prediction of the nutritional value of feedstuffs, which makes it an attractive technique for routine quality control in the industry.

Quantitative assessment of wheat quality using near‐infrared spectroscopy: A comprehensive review

“…When it comes to spectral data, there are two different approaches toward outliers. One is that outliers in spectra should be detected and removed since they may influence the performance of calibration models (Jiang et al., 2007; Shi et al., 2019; Tian et al., 2021; Viljoen et al., 2005). The most popular method adopted for spectral data outlier elimination is principal component analysis (PCA), which relies on either the F‐residuals or Hotelling's T 2 ‐values exceeding the corresponding limit of 5% (Kucheryavskiy, 2013; Shi et al., 2019; Tian et al., 2021).…”

“…The 𝑅 2 P of the two calibra-tion models was below .45 and .32, respectively, which was insufficient for quantitative determination (Dowell et al, 2006). Another study conducted by Viljoen et al (2005) had better performance by using a larger and processed dataset with 20 cultivars from 3 years as well as outliers removal (Viljoen et al, 2005). The comparison between the two studies showed that larger and processed datasets have a positive effect on model development.…”

“…Subsequently, i-PLS was introduced into ash content determination (Dong & Sun, 2013). Although Dong and Sun's (2013) sample size was half of that in Viljoen et al (2005), the performance of the PLSR model was the best and was the same as that acquired in Hell et al (2016). The ash content in wheat bran was first studied in 2016, and the study compared the performance of MIRS and NIRS (Hell et al, 2016).…”

Prediction of Triticale Grain Quality Properties, Based on Both Chemical and Indirectly Measured Reference Methods, Using Near‐Infrared Spectroscopy

“…The alternate approach of using NIR calibration models developed for wheat to predict moisture and protein contents of triticale has proven unsuccessful (Igne et al 2007b). General winter cereal NIR calibrations for protein and ash concentrations have been developed for a combined data set comprising triticale, barley, oats, and wheat samples (Viljoen et al 2005). By using triticale-specific models, these SEPs were reduced to 0.29 and 0.30% (w/w).…”

“…By using triticale-specific models, these SEPs were reduced to 0.29 and 0.30% (w/w). General winter cereal NIR calibrations for protein and ash concentrations have been developed for a combined data set comprising triticale, barley, oats, and wheat samples (Viljoen et al 2005). Ash and protein contents were quantified with SEPs of 0.22 and 0.60% (w/w), respectively, and coefficients of determination (r 2 ) of 0.76 and 0.86, respectively, for the prediction set.…”