Comparing a New Non-Invasive Vineyard Yield Estimation Approach Based on Image Analysis with Manual Sample-Based Methods

Abstract

Manual vineyard yield estimation approaches are easy to use and can provide relevant information at early stages of plant development. However, such methods are subject to spatial and temporal variability as they are sample-based and dependent on historical data. The present work aims at comparing the accuracy of a new non-invasive and multicultivar, image-based yield estimation approach with a manual method. Non-disturbed grapevine images were collected from six cultivars, at three vineyard plots in Portugal, at the very beginning of veraison, in a total of 213 images. A stepwise regression model was used to select the most appropriate set of variables to predict the yield. A combination of derived variables was obtained that included visible bunch area, estimated total bunch area, perimeter, visible berry number and bunch compactness. The model achieved an R2 = 0.86 on the validation set. The image-based yield estimates outperformed manual ones on five out of six cultivar data sets, with most estimates achieving absolute errors below 10%. Higher errors were observed on vines with denser canopies. The studied approach has the potential to be fully automated and used across whole vineyards while being able to surpass most bunch occlusions by leaves.

Computer Vision and Deep Learning for Precision Viticulture

“…As of now, models only use predictors based on detected fruits and ignore the foliage, which can be detected with a segmentation model, for instance [162]. Multiple recent works have shown the negative impact of foliage occlusion on the quality of visual yield components [65,[150][151][152].…”

“…Recent studies have shown that the fruit area might be more robust to occlusion than berry counting [150]. This work was extended by comparing yield estimation from images to the classical manual sampling approach on six parcels [151]. The authors used an artificial background and manual segmentation of the images to extract meaningful features (such as visible bunch area, canopy porosity, etc.)…”

Using NDVI, climate data and machine learning to estimate yield in the Douro wine region

“…The different approaches for vineyard yield estimation depend on the scale of implementation, and from there, direct (based on manual sampling) or indirect methods (statistical and regression models, proximal/remote sensing, and dynamic or crop simulation models) are used (Bindi et al, 1996;Sirsat et al, 2019;Taylor et al, 2019;Ubalde et al, 2007;Weiss et al, 2020). The first represent the traditional method (De La Fuente et al, 2015) susceptible to spatial and temporal variability and dependent on historical data (Victorino et al, 2022), costly and time consuming (Diago et al, 2015), with low accuracy (Tardaguila et al, 2013) and limited to small-scale application. On the other hand, indirect methods can cope with the limitations off the traditional manual sampling methods and with better results on accuracy, despite the low adoption in real commercial vineyards (Barriguinha et al, 2021).…”