Development of Partial Least Square (PLS) Prediction Model to Measure the Ripeness of Oil Palm Fresh Fruit Bunch (FFB) by Using NIR Spectroscopy

Abstract

In order to develop a model for predicting the oil palm Fresh Fruit Bunch (FFB) ripeness, a rapid and non-destructive method such as NIR spectroscopy is utilized. This method has shown its capability to determine the quality of some crops by predicting their internal chemical contents. The objective of the research is to investigate the feasibility of NIR spectroscopy to predict water and oil content in FFB by developing a calibration model. Sixty samples of FFB were scanned by using NIRFlex N-500 spectrometer ranging from 1000 to 2500nm. Water and oil content of samples were measured after scanned. To develop a calibration model, Partial Least Square (PLS) Regression and pre-processing were conducted using Unscrambler X 10.3. The results showed that PLS performs well to establish a calibration model to predict water content using MSC pre-processing with r2, factor, RSMECV, and RPD are 0.93, 3, 5.24, and 2, respectively. On the other hand, PLS could not be used well for establishing oil content calibration model because the result did not meet statistic parameters. For laboratory measurement, the model could predict water content of FFB; but it was limited to samples taken from the same variety and plantation. However, NIR Spectroscopy proposed a promising method to detect the ripeness of oil palm FFB.

Prediction of the rhodinol content in Java citronella oil using NIR spectroscopy in the initial stage developing a spectral smart sensor system - Case report

“…NIR spectroscopy can be used to analyse the components in the citronella oil fraction. These results are in agreement with (Iqbal et al 2019;Windawati et al 2019;Novianty et al 2020), which confirm that the PLS method can improve the accuracy of the NIR spectroscopy analysis to predict the various chemical compounds in materials, such as the main components in oil. This study indicated that the model to predict the rhodinol content using the PLS method could be reliable in the initial stage of the spectral smart sensor development.…”

“…This model is combined with chemometric techniques, which is a multivariate data analysis using the partial least square (PLS) regression method. PLS is a quantitative analysis method widely used in developing ideal qualities for linear analysis (Iqbal et al 2019). It also reduces the dimensions of the data The NIR transfectant spectrum data are used to develop the rhodinol levels prediction model with the GC-MS results of the chemical content information.…”

Neural network with k-fold cross validation for oil palm fruit ripeness prediction

“…The x axis is associated with the wavelength ranges of the camera detector, lens, spectrograph, and halogen lamps of 400-1000 nm region. The visible-infrared spectrum has higher intensities in the region of 700-900 nm, the same fashion found in similar research of hyperspectral and infrared imaging on oil palm fresh fruit bunches [25], [27]. The measurement using four-band optical sensors showed higher reflectance intensities in the infrared region (700-900 nm) due to fewer absorbances by chlorophyll and anthocyanin in the mesocarp layer [22].…”

“…A detection system has used a 670 nm light source and an optical sensor to determine the FFB ripeness levels [22]. Moreover, imaging techniques such as thermal imaging [23], laser-induced fluorescence imaging [24], and near-infrared (NIR) spectroscopy [25] have been proposed to determine and predict FFB ripeness.…”

A Review of AI Techniques in Fruit Detection and Classification: Analyzing Data, Features and AI Models Used in Agricultural Industry

“…A familiar example of the technology is MRI scans which are most common in hospitals. (Iqbal, Herodian, and Widodo, 2019) captured NIR spectral data with the help of NIR camera for oil palm FFB ripeness detection.…”