Preliminary assessment of hyperspectral imaging for detecting Halyomorpha halys and Nezara viridula feeding sites on bean pods: A proof-of-concept study

Abstract

Abstract This proof-of-concept study explores the potential of hyperspectral imaging (HSI) for nondestructive characterization of feeding damage by two invasive stink bug species on bean pods under controlled laboratory conditions. We compared spectral signatures of feeding sites caused by Halyomorpha halys and Nezara viridula on Phaseolus vulgaris seven days postinfestation. Using a limited dataset of 45 observations from 20 pods (12 individual plants), we identified distinct spectral modifications induced by both species. H. halys feeding increased visible spectrum reflectance (450-700 nm) with reduced near-infrared (NIR) reflectance, while N. viridula caused more severe spectral changes with substantial NIR decreases (up to 31% at 800 nm). Difference spectra revealed species-specific sensitivity patterns with maximum responses at 540 nm (green) and 740 nm (NIR) for both species, though N. viridula showed consistently larger deviations from healthy tissue. Key wavelengths for damage characterization were identified at 660-680 nm and 730-780 nm for both species. While these initial findings suggest potential spectral discrimination between stink bug feeding damage and healthy tissue, we emphasize that this preliminary study requires extensive validation with larger sample sizes, multiple cultivars, comparison with other herbivores, and field conditions before any practical application. The results provide first insights into spectral responses to stink bug feeding that warrant comprehensive follow-up investigation practical hyperspectralbased pest symptom monitoring tools that could enable more targeted management strategies against these economically significant pests.

Plant phenotyping relevance

豆莢の吸汁被害という植物状態をハイパースペクトル画像で非破壊評価する方法の概念実証であり、スペクトル特徴と識別波長の抽出が研究の中心です。

Code and data availability

The paper's hyperspectral measurements of stink bug feeding sites on bean pods and the associated R/ML analysis are not publicly deposited; the data availability statement says they are available from the corresponding author upon reasonable request.

No evidence-backed public reproduction asset is currently recorded.