Nutrient Deficiency Classification using AlexNet Architecture

Oil palm significantly contributes to Malaysia’s Gross Domestic Product (GDP). To maintain palm oil productivity, keeping the oil palm crop in good health, and free from nutrient deficiencies is necessary. Therefore, a method to analyze nutrient deficiencies is required. Nutrient deficiencies can be detected in the oil palm leaves. The study proposes an image-processing method that classifies three crucial nutrient deficiencies in a palm oil leaf: Nitrogen, Potassium, and Magnesium. To improve the processing rate, a convolution neural network (CNN) based on the AlexNet architecture has been implemented, which has been successful in image classification tasks and produced remarkable results in automated classification. The AlexNet architecture consists of five (5) convolved layers and three (3) fully connected layers, achieving an accuracy of 90.85%. This shows that AlexNet is an efficient algorithm for classifying nutrient deficiencies in palm oil leaves. However, a lower classification rate was observed for Magnesium compared to Nitrogen and Potassium. Due to the similarity in texture and color which may cause misinterpretation by the classifier. Further exploration of other CNN architectures will be conducted in the future to improve classification accuracy.