Articles
DIGIVIT: digital viticulture tool for yield and quality prediction using UAV images
Article number
1385_24
Pages
189 – 196
Language
English
Abstract
Early yield estimation and the evaluation of grape ripening plays an important role on the vineyard management and on the quality of the resulting wine.
Classical analytical procedures for ripening evaluation and yield estimation are time-consuming, expensive, and sometimes unrepresentative of vineyard variability.
Precision agriculture technologies, such as UAVs, high-resolution imaging sensors and digital image processing tools allow to avoid these problems.
This study used high-resolution RGB images from a low-cost UAV platform to estimate yield and quality parameters several weeks before harvest.
In different vineyards, four representative parcel of vigor variability zone were one-sided defoliated at fruit zone level and monitored by a low-altitude UAV with a 45° RGB camera.
An unsupervised identification method employing RGB colour filtering in HSV colour space automatically produced the segmentation of bunches, which were used to estimate production and quality.
The yield model successfully estimated the production with an R2 equal to 0.87 three weeks before harvest.
Furthermore, satisfactory calibration models were also achieved for the prediction of the most important parameters related to the grape ripening, such as anthocyanin (R2=0.81), sugar (R2=0.66) and acid malic (R2=0.56).
Classical analytical procedures for ripening evaluation and yield estimation are time-consuming, expensive, and sometimes unrepresentative of vineyard variability.
Precision agriculture technologies, such as UAVs, high-resolution imaging sensors and digital image processing tools allow to avoid these problems.
This study used high-resolution RGB images from a low-cost UAV platform to estimate yield and quality parameters several weeks before harvest.
In different vineyards, four representative parcel of vigor variability zone were one-sided defoliated at fruit zone level and monitored by a low-altitude UAV with a 45° RGB camera.
An unsupervised identification method employing RGB colour filtering in HSV colour space automatically produced the segmentation of bunches, which were used to estimate production and quality.
The yield model successfully estimated the production with an R2 equal to 0.87 three weeks before harvest.
Furthermore, satisfactory calibration models were also achieved for the prediction of the most important parameters related to the grape ripening, such as anthocyanin (R2=0.81), sugar (R2=0.66) and acid malic (R2=0.56).
Authors
A. Matese, G. Orlandi, S.F. Di Gennaro
Keywords
UAV, RGB camera, yield estimation, ripening evaluation, grapevine
Online Articles (25)