Areca Nut Disease Classification Using Sailfish Optimization Algorithm with Dynamic Elastic Boundary Strategy and Convolution Neural Networks

Abstract

In recent years, areca nut plants have been vulnerable to different diseases that appear as distinct colors on leaves, caused by bacteria or fungi. These symptoms disrupt photosynthesis and reduce yield, affecting productivity and crop health. Therefore, accurate plant disease classification is essential for detecting distinct disease shapes and sizes. Existing Deep Learning (DL) models have several limitations that prevent them from distinguishing between various plant diseases due to similar characteristics. To overcome this limitation, a Dynamic elastic boundary strategy Sailfish Optimization Algorithm and Convolution Neural Network (DSFO-CNN) method is proposed to identify and accurately classify arecanut plant diseases. The Visual Geometry Graph-19 (VGG-19) model extracts features that have significant information about disease in arecanut plants. The proposed arecanut plant disease classification model employed feature selection and drop cyclic learning rate, which adjusts the CNN learning rate to efficiently learn the subtle information about various leaf and nut diseases to enhance classification. The experimental results of the DSFO-CNN demonstrate superior performance compared to existing approaches.

Plant phenotyping relevance

アレカヤシの葉・果実に現れる病徴を画像から分類するCNNベース手法を提案・評価しており、植物病害状態の取得・推定が中心的な方法論的貢献である。

Code and data availability

The paper's phenotyping inputs are two public image datasets: the collected Arecanut dataset (Kaggle) and the PlantVillage dataset (Kaggle), both explicitly cited and declared openly available. No author analysis code or trained model is released.