Efficient early-stage disease detection in pomegranate (Punica granatum) using convolutional neural networks optimized by honey badger optimization algorithm

This study presents a novel method for early-stage disease detection in pomegranate using a convolutional neural network (CNN) and honey badger optimization algorithm (HBOA). Existing fruit disease detection methods requires the appearance of external symptoms on the fruit surface. By the time symptoms appear on the fruit surface, disease spread inside the fruit will be considerably large, it will be difficult for farmers to implement counter measures to prevent disease spread. To overcome this problem, this study presents an early-stage disease detection method for pomegranates. Initially, image quality was enhanced using contrast limited adaptive histogram equalization. Pre-processed image was segmented using k-means clustering. The features for early-stage disease detection are color-based, region-based and texture-based. The segmented image was subjected to feature extraction based on the identified features. The CNN is modified in which the extracted image features are given as input, and the modified CNN classifies the pomegranate into healthy and early-stage disease affected fruit. Classification accuracy was enhanced using the HBOA algorithm. The proposed CNN-HBOA model achieved a classification accuracy of 93.29%. To test the superiority of CNN-HBOA, early-stage disease detection was performed with existing state-of-the-art classifiers. The proposed CNN-HBOA outperformed existing classifiers with better classification accuracy.