An Enhanced Oppositional Crow Search Optimization Algorithm-based Colour Edge Segmentation and Modified Resnet-39 Architecture for Prediction of Crop Disease

Abstract

Accurate and timely prediction of crop diseases is vital for improving agricultural productivity and ensuring food security. This research work proposes new framework that integrates an improved swarm-based color edge segmentation method with a modified ResNet-39 architecture to efficiently detect and categorize crop diseases. The suggested segmentation technique utilizes Crow Search optimization algorithm combined with oppositional learning to improve edge recognition accuracy using color segmentation. This allows precise localization of disease-affected areas even under difficult situations such as poor contrast and noise. The segmented outputs are further processed using a tailored ResNet-39 architecture, optimized for computational efficiency and prediction accuracy. The proposed Enhanced edge segmentation Crow Search Optimization with ResNet-39 (EesCSO-ResNet-39) methodology is tested using publically accessible crop disease datasets, showing substantial improvements in critical metrics such as accuracy, precision, and recall relative to conventional segmentation techniques and current deep learning architectures, including ResNet-50 and VGG-16.