A Machine Learning Based Framework for a Stage-Wise Classification of Date Palm White Scale Disease

Date palm cultivation plays a vital role in the economy and ecosystem of oasis agriculture, yet it faces serious threats from diseases such as White Scale Disease (WSD) caused by Parlatoria blanchardi. This pest severely affects the health of date palms, leading to significant yield loss and, in severe cases, tree mortality. Accurate, stage-wise classification of WSD is essential for timely intervention and treatment. This study proposes a machine learning-based framework for the automatic classification of WSD stages using leaflet images. The proposed system utilizes two feature extraction techniques: 80 Gray-Level Co-occurrence Matrix (GLCM) texture features and 9 HSV color moment features. These features are extracted from grayscale and color images, respectively. Multiple classification algorithms, including Support Vector Machine (SVM), k-Nearest Neighbors (KNN), Random Forest (RF), LightGBM, XGBoost, and Multi-Layer Perceptron (MLP), are evaluated alongside a proposed stacking ensemble model. The classification targets four stages of WSD: healthy, low, medium, and high infestation. Experimental results show that combining GLCM and HSV features significantly improves performance, with the SVM classifier achieving the highest accuracy of 98.29%. The proposed stacking model further enhances reliability. This framework demonstrates the potential of intelligent feature fusion and machine learning for precise, early detection of date palm WSD.

Keywords: Date Palm, White Scale Disease (WSD), Gray-Level Co-occurrence Matrix (GLCM), HSV Color Moments, Feature Extraction, Machine Learning, Support Vector Machine (SVM), Random Forest (RF), Stacking Ensemble, Leaflet Image Classification, Plant Disease Detection, Multiclass Classification, Agricultural AI.