IBSHC: Improved BIRCH‐based segmentation and hybrid classification model for Mango Disease Detection

Abstract

Mangoes are referred to as the ‘king of fruits’ and are in great demand; yet, in order to optimize profitability, disease control is crucial. Automatic leaf disease segmentation and detection remains challenge due to the variability of symptoms. The basic necessity for any computer‐aided system is an accurate approach for identifying diseases. This paper intends to propose a four‐stage Mango Disease Detection model, IBSHC (Improved BIRCH‐based Segmentation and Hybrid Classification model for mango disease classification) that covers preprocessing, segmentation, feature extraction, and classification to address this detection problem. Initially, preprocessing phase is performed by using the Gaussian filtering process. Then the improved Balanced Iterative Reducing and Clustering using Hierarchies (BIRCH) process‐based segmentation is carried out in the segmentation phase. Subsequently, the Colour histogram, shape index histogram, improved significant Local Binary Patterns (LBP), and Resnet are the features extracted during the feature extraction phase. Finally, a hybrid classification model is developed for identifying the healthy and diseased mangoes. The optimized Dilated Residual Networks (DRN) and optimized Long Short‐Term Memory (LSTM) classifiers are employed in this hybrid model, and the PMRFO (Proposed Manta‐Ray‐Foraging Optimization) algorithm is used to optimize the weights of these two classifiers. Finally, the efficiency of the suggested PMRFO is contrasted over other conventional methods, and also the disease detection of the proposed model has better accuracy in each learning percentage like 60th LP = 92.5%, 70th LP = 93%, 80th LP = 94.02%, and 90th LP = 96.5%, respectively.