Giesco
HYPERSPECTRAL IMAGING FOR LEAF GRAPEVINE WATER STATUS ASSESSMENT, A LEAF-BASED APPROACH USO DE IMÁGENES HIPERESPECTRALES PARA LA EVALUACIÓN DE ESTADO HÍDRICO DE VIDES, ANÁLISIS A NIVEL DE HOJA
BELLOTTO, Alessandro1; MANZAN, Nicola1; GREYLING, Berno2; PITACCO, Andrea3; PETERLUNGER, Enrico1; STREVER, Albert2; POBLETE-ECHEVERRÍA, Carlos2* 1 Dipartimento di Scienze AgroAlimentari, Ambientali e Animali, University of Udine, Via delle Scienze 208, Udine, Italy Department of Viticulture and Oenology, Faculty of AgriSciences, Stellenbosch University, Matieland 7602, South Africa 3 Department of Agronomy, Food, Natural resources, Animals and Environment(DAFNAE), University of Padova Agripolis, 35020 Legnaro, Italy. 2 *Corresponding author: cpe@sun.ac.za
Abstract: Physiological measurements are the most accurate way of assessing plant water status. However, they are time-consuming, costly and invasive for the plant. Visible to shortwave infrared imaging spectrometers can be used to detect and monitor several biochemical variations in the leaf offering a faster, cheaper and non-destructive alternative to the traditional methods. In this study, hyperspectral images were used to obtain the spectral signatures of grapevine leaves (Vitis Vinifera L. cv. Shiraz) under different water conditions. Hyperspectral images were collected using a line scan sensor with a resolution of 813 x 456 pixels in the visible and near-infrared region, from 450 to 980 nm. The extraction and interpolation of the hyperspectral signatures were carried out using a customize MATLAB code. Leaf stomatal conductance (gs), measured with an Infra-red gas analyzer was used as a reference indicator of the water status. Five of most common Vegetation indexes (VIs) were calculated from spectral signatures and correlated with gs values. The preliminary results show that the linear regressions VIs-gs were not significant. However, using machine learning techniques, the information from hyperspectral imaging allow discriminating among different stomatal conductance classes, with a high accuracy especially in the images acquired around midday.
Keywords: stomatal conductance, vegetation indexes, non-destructive method, spectral signature.
Abstract: Physiological measurements are the most accurate way of assessing plant water status. However, they are time-consuming, costly and invasive for the plant. Visible to shortwave infrared imaging spectrometers can be used to detect and monitor several biochemical variations in the leaf offering a faster, cheaper and non-destructive alternative to the traditional methods. In this study, hyperspectral images were used to obtain the spectral signatures of grapevine leaves (Vitis Vinifera L. cv. Shiraz) under different water conditions. Hyperspectral images were collected using a line scan sensor with a resolution of 813 x 456 pixels in the visible and near-infrared region, from 450 to 980 nm. The extraction and interpolation of the hyperspectral signatures were carried out using a customize MATLAB code. Leaf stomatal conductance (gs), measured with an Infra-red gas analyzer was used as a reference indicator of the water status. Five of most common Vegetation indexes (VIs) were calculated from spectral signatures and correlated with gs values. The preliminary results show that the linear regressions VIs-gs were not significant. However, using machine learning techniques, the information from hyperspectral imaging allow discriminating among different stomatal conductance classes, with a high accuracy especially in the images acquired around midday.
Keywords: stomatal conductance, vegetation indexes, non-destructive method, spectral signature.