A Rule-Based Approach for Oil Palm Fruit Ripeness Detection using Machine Vision

In the oil palm industry, the grading process of oil palm fruits is conducted manually by trained inspectors via visual examination. Laborious and prone to human error, this process results in fruit ripeness being determined subjectively based on some developed standards. This study presents the development of an automatic grading system for oil palm fruit ripeness using red-green-blue (RGB) channel and rule-based classification. The grading system consists of four main stages: (i) image acquisition using camera and computer; (ii) image processing involving the segmentation of oil palm fruit; (iii) calculation of mean colour intensity based on the RGB colour model; and (iv) determination of oil palm fruit ripeness using the rule-based classification. Several features are extracted from the RGB channel and based on these extracted features; simple classification rules are formed to identify fruit ripeness. The maturity of oil palm fruits is then classified into two categories, ripe and unripe. Thirty-nine samples of oil palm fruits, comprising 21 ripe fruits and 18 unripe ones, are used to develop the classification rule. A graphical user interface (GUI) is developed using Matlab software to assist with the grading system whereby all the relevant image processing steps are coded into the GUI. The validity of the grading system is tested using nine samples of oil palm fruits (four ripe and five unripe) and the classification accuracy, sensitivity and specificity of 77.78 – 88.89%, 75%, and 80 – 100% respectively are achieved based on the established classification rule. Performance comparison with fuzzy logic indicates the promising potential of these simple classification rules as well.