Comparison between benchtop and ultra-compact handheld NIR spectrometers for the evaluation of apple quality

The use of NIR spectroscopy to measure internal quality attributes of fruits has been investigated extensively during the last decade. Several studies shown good results on the use of NIR spectroscopy to determine several apples quality parameters. In recent years, a new ultra-compact NIR spectrometer based on the linear variable filter (LVF) technology was introduced in the market. The aim of this study was to evaluate the potential of a new handheld ultra-compact NIR spectrometer (MicroNIR 1700, JDSU, CA, USA) based in the LVF technology, for rapid and non-destructive quality determination of apple fruit. The performance of this handheld device was compared to those obtained with a benchtop spectrometer (XDS, FOSS NIRSystems, MD, USA).

A total of 93 apple fruits were analysed for the following quality parameters: soluble solids (SS), titratable acidity (TA), firmness (F), DA meter index (DI), maturity index (MI) and Streif index (SI). Regarding spectra acquisition, four separate measurements using respectively the XDS and MicroNIR spectrometers were carried out on each individual apple at four points distributed equidistantly along the equatorial region. The measurements were performed by positioning the fruit directly on the spectrometers. The four spectra were averaged to obtain a mean spectrum for each fruit. Two spectral ranges, corresponding to 400-2500 nm (XDS-1) and 1100-1600 nm (XDS-2) respectively were considered with XDS spectra; while one range corresponding to 1100-1600 nm (µNIR) was considered with the MicroNIR spectra. Standard Normal Variate (SNV) was used as spectral preprocessing and multivariate calibration models were developed using Partial Least Squares (PLS) regression.

Regression models obtained using the XDS and MicroNIR spectra showed similar values regarding the root-mean-square error of cross-validation (RMSECV) and the ratio to prediction (RPD = standard deviation/standard error of cross validation), respectively, for each parameter. For SS (ºBrix): XDS-1: 0.56 and 3.4, XDS-2: 0.62 and 3.0, µNIR: 0.57 and 3.3. For TA (meq g malic acid/l): XDS-1: 0.83 and 2.1, XDS-2: 0.88 and 1.9, µNIR: 0.80 and 2.1. For F (kg/cm²): XDS-1: 0.58 and 1.8, XDS-2: 0.49 and 2.1, µNIR: 0.50 and 2.0. For DI: XDS-1: 0.14 and 3.2, XDS-2: 0.23 and 2.0, µNIR: 0.19 and 2.4. For MI: XDS-1: 0.98 and 2.6, XDS-2: 1.31 and 2.0, µNIR: 1.03 and 2.5. For SI: XDS-1: 0.02 and 1.5, XDS-2: 0.02 and 1.4, µNIR: 0.02 and 1.6. The prediction performance of MicroNIR appears to be similar to the benchtop spectrometer (XDS) to determine the most apples quality parameters evaluated in our current study.