Multi-classification of disease induced in plant leaf using chronological Flamingo search optimization with transfer learning

Abstract

Agriculture is imperative research in visual detection through computers. Here, the disease in plants can distress the quality and cultivation of farming. Earlier detection of disease lessens economic losses and provides better crop yield. Detection of disease from crops manually is an expensive and time-consuming task. A new scheme is devised for accomplishing multi-classification of disease using plant leaf images considering the chronological Flamingo search algorithm (CFSA) with transfer learning (TL). The leaf image undergoes pre-processing using Adaptive Anisotropic diffusion to discard noise. Here, the segmentation of plant leaf is done with U-Net++, and trained by the Moving Gorilla Remora algorithm (MGRA). The image augmentation is further applied considering two techniques namely position augmentation and color augmentation to reduce data dimensionality. Thereafter the feature mining is done to produce crucial features. Next, the classification in terms of the first level is considered for classifying plant type and classification in terms of the second level is done to categorize disease using convolutional neural network (CNN)-based TL with LeNet and it undergoes training using CFSA. The CFSA-TL-based CNN with LeNet provided better accuracy of 95.7%, sensitivity of 96.5% and specificity of 94.7%. Thus, this model is better for earlier plant leaf disease detection.