| Autor: |
Azmi MHIM; Department of Electrical, Electronic and Systems Engineering, Faculty of Engineering and Built Environment, Universiti Kebangsaan Malaysia, Bangi 43600, Selangor, Malaysia., Hashim FH; Department of Electrical, Electronic and Systems Engineering, Faculty of Engineering and Built Environment, Universiti Kebangsaan Malaysia, Bangi 43600, Selangor, Malaysia.; Research Centre for Sustainable Process Technology (CESPRO), Faculty of Engineering and Built Environment, Universiti Kebangsaan Malaysia, Bangi 43600, Selangor, Malaysia., Huddin AB; Department of Electrical, Electronic and Systems Engineering, Faculty of Engineering and Built Environment, Universiti Kebangsaan Malaysia, Bangi 43600, Selangor, Malaysia., Sajab MS; Research Centre for Sustainable Process Technology (CESPRO), Faculty of Engineering and Built Environment, Universiti Kebangsaan Malaysia, Bangi 43600, Selangor, Malaysia.; Department of Chemical and Process Engineering, Faculty of Engineering and Built Environment, Universiti Kebangsaan Malaysia, Bangi 43600, Selangor, Malaysia. |
| Abstrakt: |
The degree of maturity of oil palm fresh fruit bunches (FFB) at the time of harvest heavily affects oil production, which is expressed in the oil extraction rate (OER). Oil palm harvests must be harvested at their optimum maturity to maximize oil yield if a rapid, non-intrusive, and accurate method is available to determine their level of maturity. This study demonstrates the potential of implementing Raman spectroscopy for determining the maturity of oil palm fruitlets. A ripeness classification algorithm has been developed utilizing machine learning by classifying the components of organic compounds such as β-carotene, amino acid, etc. as parameters to distinguish the ripeness of fruits. In this study, 47 oil palm fruitlets spectra from three different ripeness levels-under ripe, ripe, and over ripe-were examined. To classify the oil palm fruitlets into three maturity categories, the extracted features were put to the test using 31 machine learning models. It was discovered that the Medium, Weighted KNN, and Trilayered Neural Network classifier has a maximum overall accuracy of 90.9% by using four significant features extracted from the peaks as the predictors. To conclude, the Raman spectroscopy method may offer a precise and efficient means to evaluate the maturity level of oil palm fruitlets. |
