Plant Disease Detection and Classification Based on Rat Swarm Optimization using Deep Learning Approach

Abstract

Plant diseases and pests cause considerable agricultural and ecological damage. Earlier prevention and detection of plant diseases are the most important aspect of crop harvesting since they can efficiently decrease growth disorders, thereby minimalizing the pesticide application for pollution-free crop harvesting. In that regard, automatic recognition of plant ailment using diverse Deep Learning (DL) and Machine Learning (ML) models has become an effective approach for precision agriculture. This article introduces a Rat Swarm Optimization with DL-based Plant Disease Detection and Classification (RSODL-PDDC) approach. The presented RSODL-PDDC technique is focused on the recognition and categorizing of plant diseases by implementing Computer Vision (CV) and DL models. Initially, image preprocessing is performed for noise removal that occurs in the images. Moreover, the RSODL-PDDC algorithm uses an attention based RetinaNet model for image segmentation purposes. Besides, the Class Attention Learning (CAL) layer is exploited to capture the discriminatory class-specific features while class-wise attention based on the Inception v3 (I-v3) model is applied for fine-grained semantic feature maps. For adjusting the hyper-parameters of the I-v3 method, the RSO algorithm is employed. Finally, Adaptive Neuro-Fuzzy Inference System (ANFIS) technique can be exploited for plant disease categorization. In order to validate the performance of the RSODL-PDDC technique, a widespread experimental analysis is performed. The result analysis pointed out the enhancements of the RSODL-PDDC algorithm over other approaches.