DEVELOPMENT OF AN ENHANCED HYBRID INTELLIGENT MACHINE LEARNING MODEL FOR MAIZE LEAF DISEASE DETECTION AND CLASSIFICATION

Abstract

Maize leaf diseases are common and have adverse effects on agricultural productivity by reducing crop yield and grain quality, which leads to significant economic losses and raises concerns about food security. Although Support Vector Machine (SVM) techniques have been widely applied for automated disease diagnosis, their effectiveness is often limited by inadequate optimisation, resulting in reduced classification accuracy and poor generalisation. This research presents an Enhanced Hybrid Intelligent Machine Learning Model (EHIMLM-SVM) that integrates the Enhanced Binary Particle Swarm Optimisation (EBPSO) and the Enhanced Reptile Search Algorithm (ERSA) for optimisation of the classifier. The model was evaluated using 2,222 maize leaf images obtained from the Kaggle PlantVillage dataset. Preprocessing involved grayscale conversion, contrast enhancement, adaptive median filtering, Sobel edge detection, and extraction of colour, texture, and shape features. Experimental results produced 97.50% accuracy, 98.15% precision, 97.76% sensitivity, 97.80% specificity, and a 2.20% false positive rate, indicating reliable performance for precision agriculture and sustainable crop management.