A Cloud-Based Platform for Leaf Disease Segmentation and Classification using Hybrid Deep Learning Model

Abstract

This paper presents a cloud-based hybrid deep learning method for classifying leaf diseases. At first, sensors aid in detecting soil factors such as moisture content, pH level, air temperature, and humidity. The hybrid Recurrent Layer based Deep Convolutional Neural Network (RLDCNN) approach consists of five distinct stages: preprocessing, segmentation, handcrafted feature extraction, feature selection, and classification. The preprocessing of image involves removal of noise using Averaged Double Plateau Histogram Equalisation (ADPHE). Afterwards, the specific part of the image that is impacted is divided into segments using an improved Wavelet transform. After the process of segmentation, the GLCM (Gray-Level Co-occurrence Matrix) is used to extract features. The hybrid Fire Hawk with Egret Swarm Optimisation (FHESO) algorithm is a very efficient approach for feature selection, significantly lowering the dimensionality of features. The detection of apple, banana mango and groundnut leaf disease are accomplished via the use of a Recurrent Layer based Deep Convolutional Neural Network (RLDCNN), which is trained using the Deer Hunting Algorithm (DHA) that has been created. The suggested concept is implemented inside a cloud-based collaborative framework to provide immediate assistance to farmers. The experimental findings exhibit favourable outcomes, demonstrating a high level of accuracy, precision, a low rate of loss, recall, and F-score. The RLDCNN network method outperforms previous parameters, obtaining an accuracy of 98.51% according to comparative research.