A feasibility study on non-destructive determination of oil content in palm fruits by visible-near infrared spectroscopy

Palm oil content is the most important factor for quality evaluation of the fresh oil palm fruits. Evaluating the colour of fruits with the naked eye and weighing of fresh fruit bunches are employed as the rapid method for estimating the palm oil content, but may not be accurate. On the other hand, the chemical extraction method yields more reliable results but it is a time-consuming method. Therefore, a reliable and rapid method for the determination of the oil content in oil palms is required. This study investigated the feasibility of non-destructive near infrared (NIR) spectroscopy for the non-destructive determination of palm oil in oil palm fruits. A total of 42 oil palm fruit samples, at varying stages of ripeness, were examined. The samples were scanned by two NIR spectrometers with different modes; one used a diffuse reflectance mode in the long wavelength (LW) region of 1100-2500 nm and the other employed an interactance mode in the visible-short wavelength (vis-SW) region of 665-955nm. The palm oil content was analysed by soxhlet extraction as a reference method. Partial least squares regression (PLSR) models for the determination of the palm oil content in the oil palm fruits were developed for both the vis-SW- and LW-NIR spectral regions and compared. The results showed that NIR spectroscopy can be applied for the determination of palm oil content in fresh fruits as a reliable, rapid and non-destructive method. The model developed from the LW-NIR region has modestly higher performance for correlation coefficients of determination (R-2) of 0.92, root mean square error of calibration (RMSEC) of 1.80%w/w, root mean square error of cross-validation (RMSECV) of 2.81%w/w, the ratio of standard deviation of reference data to RMSECV in the calibration set (RSC) of 2.19 than those obtained from the vis-SW NIR region.