LEAF DISEASE SEVERITY ESTIMATION IN ARABICA COFFEE USING A HYBRID DEEP LEARNING APPROACH

Abstract

Leaf disease severity estimation is an important factor of consideration when it comes to managing agricultural productivity. This approach helps in making timely interventions to mitigate crop losses and ensure food security. This research introduces a novel approach for the accurate assessment of disease severity in Arabica coffee leaves by integrating a novel handcrafted feature descriptor with deep learning. This approach leverages the feature extraction capabilities of a pre-trained ResNet50 convolutional neural network and complements them with the handcrafted features from a Multi-trend Binary Code Descriptor (MTBC). This hybrid strategy generates an enriched and discriminative feature set that effectively captures both high-level and textural characteristics of diseased leaves. The combined feature vector is used to train machine learning classifiers. Through extensive experimentation, the Random Forest classifier achieved the highest performance with accuracy of 91.72% on Leaf dataset and 97.85% on Symptom dataset. This work contributes to the field of precision agriculture by providing a reliable and data-driven tool for the early and precise estimation of disease severity. It is anticipated that this approach will facilitate proactive disease management to enhance crop health monitoring in Arabica coffee cultivation and other vital crops.