Assessing wine grape quality parameters using plant traits derived from physical model inversion of hyperspectral imagery

Assessing Grapevine Biophysical Parameters From Unmanned Aerial Vehicles Hyperspectral Imagery

“…For instance, floral bud induction for next season cropping is typically decided in grapevine the season before the image is taken and it is controlled, among several factors, by specific growth and environmental conditions at that time ( May, 2004 ). As per final grape composition, the overall mild correlations found for all VIs vs. TSS (°Brix) at harvest support recent work by Suarez et al (2021) , who also reported non-significant correlations for data taken on Shiraz. It is quite notable from our study that a few VIs which already provided good correlations with Ψ pd (EVIm, SR680, RVI), were those also having close correlation with malic acid concentration at harvest ( R 2 = 0.61–0.63) as well as with total anthocyanins and phenols concentration ( R 2 = 0.40–0.48).…”

“…Such changes might be a function of the inherent vine vigor, type of training system, timing and extent of leaf removal or shoot thinning ( Poni et al, 2018 ); (ii) taking RVI as the best example, it is viticulturally quite relevant and useful to have HS indices warranting good correlation with a range of variables representing plant water status (e.g., Ψ pd ), crop potential (e.g., fresh berry weight), and degree of maturity (e.g., TSS, malate, anthocyanins, and phenols). We also feel that this is the first time this achievement is reported, as in the Suarez et al (2021) paper the very close correlation that several HS indices show with a few terpene compounds is not reflected in any significant correlation with either color and phenolics.…”

“…Furthermore, the results of this study showed better performance than those obtained by Suarez et al (2021) when using plant traits derived from physical model inversion of hyperspectral imagery for the evaluation of qualitative parameters of grapevine, such as phenolic content. Considering that the grapevine is a complex system characterized by a dynamic balance between vegetative and productive features, another strong point of this work is proving the potential of a hyperspectral imaging sensor on the main key factors of the “vine-ecosystem.” In fact, compared to other cited works, a wide scenario has been explored, both functional aspects related to the eco-physiological state, as well as the vegetative growth and finally the quantitative and qualitative productive response at the end of the cycle.…”

“…These methods modify the spectral feature space so that the resultant (latent) factors account for the most variation in the feature space (PCA), or in the covariance with the target variables (PLS). State-of-the-art research presents only 7 papers encompassing UAV-based HS applications in a vineyard ( Zarco-Tejada et al, 2012 , 2013 ; Vanegas et al, 2018 ; Horstrand et al, 2019 ; Maimaitiyiming et al, 2020 ; Suarez et al, 2021 ; Di Gennaro et al, 2022 ), while more than 53 papers focused on HS applications in a vineyard without the use of UAVs. Di Gennaro et al (2022) suggested a comparison in term of accuracy between broadband multispectral and narrowband HS data by means the calculation of some VIs on canopy and soil targets in vineyard, assessing in general higher spectral accuracy of HS camera respect to the ground truth provided by reference spectroradiometer ( Di Gennaro et al, 2022 ).…”

“… Di Gennaro et al (2022) suggested a comparison in term of accuracy between broadband multispectral and narrowband HS data by means the calculation of some VIs on canopy and soil targets in vineyard, assessing in general higher spectral accuracy of HS camera respect to the ground truth provided by reference spectroradiometer ( Di Gennaro et al, 2022 ). Suarez et al (2021) investigated the links between grape quality parameters such as aroma components vs. image-based spectral indices and photosynthetic plant traits derived by physical model inversion methods. Maimaitiyiming et al (2020) considered aerial hyperspectral and thermal images acquired by using a visible and near-infrared (VNIR, 400–1,000 nm) push-broom hyperspectral camera (Nano-Hyperspec VNIR model, Headwall Photonics, Fitchburg, MA, United States) installed in tandem with a thermal camera (FLIR Vue Pro R 640, FLIR Systems, Inc., Wilsonville, OR, United States) carried by a hexacopter (Matrice 600 Pro, DJI Technology Co., Ltd., Shenzhen, China).…”