CORRECT AND INCORRECT PALM OIL FERTILIZING ACTIVITY RECOGNITION USING SMARTPHONE SENSOR AND ARTIFICIAL NEURAL NETWORK

Abstract

A good management of fertilizing activity is very important in ensuring excellent growth and yield of a palm oil farms. The advancement in sensor and machine learning technology allows fertilizing activity to be modelled and intelligent classification decision can be obtained from the model. This work proposes the use of smartphone sensors and artificial neural network for the classification of fertilizing activity either it follows standard operating procedures or not. The dataset for the proposed system is developed from the smartphone that has been attached to the body of the subjects applying actual fertilizer at palm oil farm. Mean, variance, kurtosis, zero crossing, root mean square, mean absolute deviation and inter-quartile range feature extraction method are applied on the dataset to transform the raw time series signal into a meaningful characteristic before feeding into the artificial neural network. The fully optimized model has produced the highest 81.81% correct classification accuracy rate. Hence this system is very reliable to help the palm oil planter monitoring the workers fertilizing activity.