Enhanced Disease Detection in Pomegranate Cultivation Using PF-CNN: A Deep Learning Approach for Improved Yield and Quality Management

Abstract

Fruit cultivation is a key contributor to the agricultural economy, with pomegranate being particularly valued for its high nutritional content, including antioxidants, vitamins, and fiber. Pomegranate crops are frequently susceptible to various diseases, which can lead to a significant reduction in both yield and quality. In this study, we propose a novel method for disease detection in pomegranates, utilizing a Pooling_mean/max and Filtering_Median technique based on the Convolutional Neural Network (PF-CNN). The methodology integrates median filtering with Pooling_mean/max, followed by relevant extraction of features through transfer learning from a pretrained ResNet101 architecture. The fruit images are segmented into 3x3 grids, with each segment subjected to median filtering and pooling operations before being recombined. The PF-CNN model's performance is assessed using a Support Vector Machine (SVM) classifier. Research outcomes reveal a robust classification accuracy of 96.21% across five distinct categories, including various diseases and healthy states of pomegranate fruit. This model facilitates the timely identification of diseases, enabling users to implement appropriate interventions and improve disease management strategies in pomegranate cultivation.