Use of machine learning techniques to detect and classify selected fungal diseases in rice crop using hyperspectral imaging

Abstract

Fungal diseases cause significant yield losses in rice, making early detection and accurate classification essential for effective disease management. In this study, hyperspectral imaging technique was used to acquire the spectral signatures of three major fungal diseases viz., brown spot, blast and sheath blight in rice. The acquired hyperspectral images were pre-processed using Standard Normal Variate (SNV) transformation and Savitzky-Golay filtering, followed by pixel-wise spectral data extraction. Principal Component Analysis (PCA) was used to investigate spectral variability among healthy and diseased leaf samples. Subsequently, machine learning models including artificial neural networks (ANN), support vector machines (SVM) and random forests (RF) were employed to classify these diseases based on the acquired and pre-processed spectral signature data. The results indicated that the ANN model outperform the others, achieving an accuracy of 98%, followed by SVM at 94%, and RF at 88%. Among the three models, the ANN exhibited the highest accuracy, precision and recall, making it the most effective model for disease detection and classification. Hyperspectral imaging, combined with machine learning, offers an affordable and efficient solution for large-scale detection and assessment of fungal diseases in rice crops.

Plant phenotyping relevance

イネ葉の病害状態をハイパースペクトル画像から取得し、機械学習で検出・分類する手法が研究の中心であり、植物病害表現型の技術評価に該当する。

Code and data availability

公開論文であることは確認できましたが、現在の公式API・許可済み取得経路では本文を自動取得できませんでした。

No evidence-backed public reproduction asset is currently recorded.