Hyperspectral image extraction to evaluate the photosynthetic and stress status of plants, using a photochemical reflectance index (PRI)
Tetsu Ogawa, Maro Tamaki, Takae Usui, Kouki Hikosaka
Abstract
• PRI is useful to detect plant stress, but is disturbed by noise sources in images. • We developed a new method to extract plant stress from hyperspectral images. • Our method eliminates non-leaf objects, edges, specular reflection, etc. • Extracted PRI values successfully distinguish stressed and non-stressed plants. Environmental stressors such as drought and heat often reduce plant productivity and yield. Early identification of stress responses can help to avoid yield losses. Although the photochemical reflectance index (PRI) has been widely investigated as a detection method for stress response, it is influenced not only by the stress status of the plant, but also by other factors, such as signals from non-leaf objects like soil and stems, self-shading of leaves, and specular reflections. Thus, PRI may not be reliable unless these confounding effects are addressed. In this study, we developed a new method to extract plant stress status information from high-resolution spectroscopic images. This new system filters pixels less affected by non-leaf objects, shaded leaves, and specular reflections. Additionally, it selects leaf pixels exposed to similar irradiances, which is crucial because PRI is a light-dependent variable. This new filtering method successfully distinguished drought stress in tomato plants. Thus, a robust framework for evaluating stress responses in plant production has been developed.