An Efficient Optimization Technique for Classification of Multi-Crop Leaf Diseases Using Hybrid Deep Learning Model

Abstract

Plant diseases are undesirable factors that drastically lower crop quality and production. The risk that a plant will sustain additional damage may be reduced by early and precise detection of plant diseases. This research recommends a new hybrid model called BOA-CNN-LSTM that can autonomously identify and categorize multi-class leaf diseases with the highest detection rate. The provided approach pre-processes the data to extract features and boost the classification accuracy readily. The CNN technique is used first to extract the features from the plant leaf images. The LSTM technique is then applied to classify each image. Additionally, the CNN-LSTM model's hyper-parameter optimization utilizes the butterfly optimization algorithm (BOA), which improves detection performance. The plant village database is utilized to evaluate the functionality of the proposed approach, and the results are examined in terms of multiple assessment criteria. Four distinct plant types—apples, corn, tomatoes, and rice—are employed in the test. The proposed work achieves an accuracy of 92%, sensitivity of 0.91, and specificity of 0.89, according to the experimental findings of the study.