A compact multimodal and imaging system for presymptomatic plant stress detection in NASA-controlled space agriculture

Abstract

A hybrid AI framework combining a spatial–spectral–temporal Transformer and unsupervised clustering was applied to five microgreen species—Pak Choi Cabbage, Tatsoi Mustard, Red Mizuna, Chinese Cabbage, and Arugula—grown for 3–4 weeks under water, nutrient, and combined stresses. Across five datasets collected within six months, the system achieved a Macro-F1 of 0.91 and a pre-symptomatic F1 of 0.88, enabling early detection before visible symptoms. Applications include NASA’s APH, Mars and Moon habitats, and terrestrial precision agriculture.

Plant phenotyping relevance

植物の水・養分ストレス状態をマルチモーダル画像から早期推定するAI手法が研究の中心であり、性能評価も示されている。

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.