An Ensemble Hybrid Framework: A Comparative Analysis of Metaheuristic Algorithms for Ensemble Hybrid CNN Features for Plants Disease Classification

A stable and sustainable food supply chain is only possible with effective agriculture management. The application of technology in agriculture has recently produced encouraging outcomes in terms of improving agricultural yield and quality. Early detection and characterization of plant leaf diseases, which have a substantial influence on agricultural output and the general well-being of farmers and consumers, is a crucial component. The automatic identification and categorization of illnesses affecting apple and maize plants is the focus of this paper's investigation into the implementation of cutting-edge technical solutions, notably machine learning algorithms. Deep learning has recently made significant contributions to the automatic detection and classification of plants diseases specifically in fruits and vegetables. As a result, the production and quality of fruits and vegetables have both improved. Many diseases have a negative impact on crop quality and output. This research proposes an ensemble deep learning based technique for classifying plant leaf diseases. In this work, two separate pre-trained deep learning models have been utilized and a hybrid approach is proposed. In this research work we have proposed a hybrid framework based on the hybrid preprocessing algorithm, an ensemble features engineering phase based on the texture features and two types of deep features extraction. The extracted Convolutional Neural Network (CNN) features are selected and fused with the LBP features. The ensemble feature vector is optimized using three different meta-heuristic algorithms that are Binary Dragonfly algorithm (BDA), Ant Colony Optimization algorithm and Moth Flame Optimization algorithm (MFO). The optimized feature vector is classified using state-of-the-art machine learning algorithms. Notably, the study addresses the research gap by providing a comprehensive analysis of plant diseases affecting fruits and vegetables. Additionally, it acknowledges the limitations of the proposed approach, paving the way for future research and improvements. The utilization of the proposed technique yields a remarkable accuracy of 99.8%, surpassing current state-of-the-art methods. This research contributes to the ongoing efforts to enhance food security and sustainability by providing an effective solution for the early identification and classification of plant diseases, thereby mitigating the negative impact on crop productivity and quality.