Enhanced Leaf Disease Diagnosis in Black‐Pepper Plants Using a Dense Back‐Projection Joint Attention Network

Agriculture is a fundamental sector that ensures food security, particularly in the context of a rapidly growing global population. In this manuscript, Enhanced Leaf Disease Diagnosis in Black‐Pepper Plants Using a Dense Back‐projection Joint Attention Network (DBJAN‐LDD‐BPP) is proposed. Initially input images are gathered from plant village + Black Pepper Leaf Blight and Yellow Mottle Virus dataset. Preprocessing involves two stages: Local Sample‐weighted Multiple Kernel Clustering (LSMKC), it identifies infected regions in leaves and Adaptive Bitonic Filtering (ABF) performs data augmentation to increase dataset diversity. Then, pre‐processed images are feature extracted utilising Multi‐Objective Matched Synchrosqueezing Chirplet Transform (MOMSCT), to extract texture features. Then, extracted images are fed to Dense Back‐Projection Joint Attention Network (DBPJAN) for effectively categorise black‐pepper leaf disease Detection as Pepper bell Bacterial spot, Pepper bell healthy, black pepper leaf blight and black pepper yellow mottle virus. Hence, the Wild Horse Optimization Algorithm (WHOA) to optimise DBPJAN which accurately classifies the Black‐Pepper Leaf Disease. Then proposed DBJAN‐LDD‐BPP is implemented and performance metrics like accuracy, precision, sensitivity, specificity, computation time and ROC are examined. The efficiency of theDBJAN‐LDD‐BPPachieves 14.92%, 25.75% and 32.18% higher accuracy when compared with existing YOLOv5‐LDD‐BPP, CNN‐LDD‐BPP and SE‐VRNet LDD methods, respectively.