Application of aerial remote sensing technology for detection of fire blight infected pear trees

Abstract

Over the last few decades, Fire Blight (FB) has recognized as the most dangerous diseases of apple and pear trees in the world. Timely diagnosis is very important for the detection of this disease. Visual assessment and scouting are usually used for FB detection while these methods are time-consuming and labor-intensive. Remote sensing technology can be an alternative method for visual assessment of plant diseases. So, in this research, the capability of multispectral remote sensing was evaluated for FB disease diagnosis of pear orchards in leaf and tree crown levels. Ground multispectral imaging was carried out of healthy leaves (HEL) from healthy trees and non-symptomatic diseased leaves (NSL) and symptomatic diseased leaves (SDL. Aerial multispectral imagery of trees crown was carried out by unmanned aerial vehicle. Then preprocessing and processing of ground and aerial images were performed. Some vegetation indices were calculated to detect infected leaves. Among the studied indices, SIPI, RDVI, MCARI1, MCARI2, TVI, MTVI1, MTVI2, TCARI, PSRI and ARI indices were appropriate for early detection of FB in leaf level. Support vector machine (SVM) method was used for the detection of infected trees. The overall accuracy of the classification obtained 95.0%. Based on the results, it could be concluded that multispectral imaging in leaf and tree crown levels is a reliable method for the detection of FB infected pear trees spatially in the early stage.

Plant phenotyping relevance

ナシ樹の火傷病症状をマルチスペクトル画像とSVMで検出・分類する手法を開発・評価しており、植物病害状態の取得が中心である。

Code and data availability

公開本文の所在を確認できませんでした。非公開または購読が必要な可能性があります。

No evidence-backed public reproduction asset is currently recorded.