Performance of two handheld NIR spectrometers to quantify crude protein of composite animal forage and feedstuff

<abstract> <p>Two handheld near infrared (NIR) spectrometers were used to quantify crude protein ($CP$) content of mixed forage and feedstuff composed Sweet Bran, distiller's grains, corn silage, stalk. First was a transportable spectrometer, which measured in the visible NIR ranges (320–2500 nm) with spectral interval 1 nm (H1). Second smartphone from 900–1700 4 (H2). Spectral data 147 feed samples collected by both instruments split into calibration ($n$ = 120) validation 27) sets. For H1, only absorbances region (780–2500 multivariate analyses, while for H2, second third overtone regions (940–1660 used. Principal component analysis (PCA) partial least squares (PLS) regression models developed using mean-centered that had been preprocessed standard normal variate (SNV) or Savitzky-Golay first derivative (SG1) (SG2) algorithm. PCA showed two major groups—one Bran distillers other silage Using H1 spectra, PLS model best predicted $CP$ followed SG1 preprocessing. This low root mean square error prediction ($RMSEP$ 2.22%) high ratio deviation ($RPD$ 5.24). With H2 predicting based on SG2 preprocessing, returning $RMSEP$ 2.05% $RPD$ 5.74. These values not practically different than those indicating similar performance devices despite having absorbance measurements H2. The result this study can accurately measure during screening, quality, process control applications.</p> </abstract>