Open Access Open Access  Restricted Access Subscription Access

Copyright © 2023 ISEIS. All rights reserved

Rice Plant Leaf Disease Detection and Classification Using Optimization Enabled Deep Learning

T. Daniya1,2 * and S. Vigneshwari1

  1. Department of Computer Science and Engineering, Sathyabama Institute of Science and Technology, Chennai 600119, Tamilnadu, India
  2. Department of Information Technology, GMR Institute of Technology, Rajam 532127, Andhra Pradesh, India

*Corresponding author. Tel.: +91 9095051376; fax: 08941-251591. E-mail address: (T. Daniya).


An automatic identification and classification of rice diseases are very important in the domain of agriculture. Deep learning (DL) is an effective research area in the identification of agriculture pattern identification where it can effectively resolve the issues of diseases identification. In this paper, a hybrid optimization algorithm is developed to categorize the plant diseases. The pre-processing is made using Region of Interest (ROI) extraction and the input image is created by combining the Rice plant dataset, and Rice disease dataset. The segmentation is accomplished using Deep fuzzy clustering. The features, like statistical features, entropy, Convolutional Neural Network (CNN) features, Local Optimal-Oriented Pattern (LOOP), and Local Gabor XOR Pattern (LGXP) is considered for extracting the appropriate features for further processing. The data augmentation is employed to enlarge the volume of extracted features. Then, the first level classification is made by deep neuro-fuzzy network (DNFN), which is trained using Rider Henry Gas Solubility Optimization (RHGSO) that categories into healthy and unhealthy plants. The RHGSO is the integration of Rider Optimization Algorithm (ROA) and Henry gas solubility optimization (HGSO). After that, second-level classification is made by a Deep residual network (DRN) that is tuned by RHGSO. Thus, the RHGSO-based DRN categorizes the unhealthy plants into Bacterial Leaf Blight (BLB), Blast, and Brown spot. Thus, the implementation of the proposed RHGSO-based deep learning approach offered better accuracy, sensitivity, specificity, and F1-score of 0.9304, 0.9459, 0.8383, and 0.9142.

Keywords: deep residual network, texture features, neural network, henry gas solubility optimization, local optimal-oriented pattern

Full Text:


Supplementary Files:


  • There are currently no refbacks.