ADAPTIVE HYBRID FRAMEWORK FOR ROBUST APPLE LEAF DISEASE DETECTION USING AHE AND GMM

Apple leaf diseases pose a significant threat to global apple production, demanding accurate and efficient diagnostic frameworks.This study proposes a hybrid model integrating advanced pre-processing techniques, feature extraction, and robust classification methodologies to enhance disease detection accuracy.Pre-processing techniques such as Adaptive Histogram Equalization (AHE) and Gaussian Mixture Models (GMM) were employed to improve image quality and segment disease regions effectively.Feature extraction was performed using a pre-trained ResNet-50 model, followed by classification using a CNN-SVM hybrid model, enhanced with an ensemble Random Forest classifier for robustness.Evaluated on the Plant Village dataset, the proposed model achieved a mean accuracy of 97.76%, outperforming existing methods.This study underscores the importance of advanced pre-processing and hybrid modelling in precision agriculture, offering a scalable and accurate solution for early disease detection and sustainable apple farming practices.