Oil content determination scheme of postharvest oil palm for mobile devices

In this paper, a simple scheme based on an image processing technique for determining oil content in postharvest oil palm (Elaeis guineensis Jacq.), implemented for a mobile device, is proposed. The scheme has three main algorithms for colour correction, classification and oil extraction rate (OER) determination. The colour correction algorithm can correct image colour from the device-dependence effect in the standard RGB (sRGB) colour model based on the determined device profile function. The classification process was developed on a two-layer feedforward neural network by using features from the hue value of oil palm fruits. The OER determination function was modelled by using a polynomial regression model based on the hue and saturation values. The results demonstrated that the proposed scheme can classify and determine the OER with a simple calculation. The scheme was implemented on a mobile device/phone and tested with 64 oil palm fruit samples. Compared with the standard Soxhlet extraction measurement, the scheme achieves a mean error of OER 2.20 with a postharvest OER range of 30-73%.