Application of Artificial Neural Network Classification to Determine Nutrient Content in Oil Palm Leaves

Abstract

Abstract. The classification of oil palm nutrient level based on leaf samples is an important factor to dictate the quality of fresh fruit bunch (FFB). The optimum nutrient level in a palm tree ensures high yield and productivity. This study evaluated a spectroradiometer of spectral bands ranging from 350 to 2500 nm to detect nutrient level in oil palm leaf samples. The features considered were types of nutrient and fronds, explored in spectral reflectance of wavelength for nutrient level determination. Results from statistical analysis using the spectral reflectance of oil palm leaves with partial least square (PLS) models were used for classification of three nutrient levels, comprising of low, optimum, and high amount of fertilization, using the artificial neural network (ANN) to inspect oil palm leaves for contents of nitrogen (N) and potassium (K). From the 90 leaf samples, the ANN models had classification performance of 85.32% accuracy for oil palm nutrient contents determination and 69.42% accuracy for frond identification. Results of this study imply the use of ANN as a prime tool for classification and identification of features in oil palm leaves. Keywords: Artificial neural network (ANN), Oil palm, Nutrients, Spectroradiometer.