Intelligent Optimization and Deep Learning Approaches to Pomegranate Disease Detection

• Mayilai Thanikesan Vasumathi,
• Mari Kamarasan,
• Balasubramanian Karthikeyan,
• Ganesan Manikandan,

Vol. 19, No. 7, July 31, 2025
10.3837/tiis.2025.07.005, Download Paper (Free):

• active learning
• Butterfly Optimization Model
• Classification
• Deep Learning
• Deep Belief Network

Abstract

Pomegranates have multiple health benefits and unique taste, making them popular worldwide. However, the productivity of pomegranates is severely affected in terms of quality and quantity by fruit and leaf diseases. Automated plant disease detection has become a major challenge for smart farming, attracting the highest level of attention among research communities. The rise of deep learning (DL) concepts is found to be advantageous in the diagnosis and prediction of crop diseases. Given this standpoint, this paper introduces a novel Butterfly Optimization Algorithm-based Deep Belief Network (BOA-DBN) model for automated pomegranate fruit disease detection and classification. The BOA-DBN model aims to detect and classify four fruit diseases with a maximum detection rate. The presented model initially performs a data acquisition process to gather the required pomegranate fruit images and then record the features in an Excel data format. Next, data normalization is conducted to improve the detection accuracy rate. Specifically, the DBN model is applied as a classifier for pomegranate fruit disease detection, and BOA is used for the hyperparameter optimization of the DBN model. Overall, the optimization improves the detection performance. The performance of the BOA-DBN model was assessed using over 6000 records, and the results were examined in terms of different evaluation metrics. Compared to other methods, the BOA-DBN model achieved superior results, including a maximum accuracy, precision, recall, specificity, and F-score of 0.9930, 0.9800, 0.9898, 0.9984, and 0.9848 respectively.

Statistics

Cite this article