Using image analysis for assessing downy mildew severity in grapevine
Abstract
Aim: Downy mildew is a crucial disease in viticulture. In-field evaluation of downy mildew has been classically based on visual inspection of leaves and fruit. Nevertheless, non-invasive sensing technologies could be used for disease detection in grapevine. The aim of this study was to assess downy mildew severity in grapevine leaves using machine vision.
Methods and Results: Leaf disks of the cv Pinot Noir (Vitis vinifera L.) were placed in Petri dishes with the abaxial side up. Plasmopara viticola sporangia were collected from infected leaves in the vineyard and used for the experimental inoculation of the leaf disks in laboratory. Images of Petri dishes including different levels of downy mildew infection were taken using a digital RGB camera. Machine vision techniques were used to estimate downy mildew severity (percentage of pixels representing visual symptoms) on the leaves. The symptoms were evaluated by eight experts, visually estimating the percentage of area showing sporulation. Considering the average evaluation of the experts, the assessment obtained by the new developed algorithm based on computer vision was represented as a R2value of 0.82 and RMSE of 14.34%.
Conclusions:
The results show a strong correlation between the severity computed by machine vision and the visual assessments, opening the possibility of the automated evaluation of downy mildew severity using non-invasive sensors.
Significance and Impact of the Study: The results indicated that machine vision can be applied for assessing and quantify visual symptoms of downy mildew in grapevine
DOI:
Issue: Terroir 2020
Type: Video
Authors
1Televitis Research Group. University of La Rioja, 26007 Logroño, Spain
2Department of Computer Science and Engineering, University of Cádiz, 11519 Puerto Real, Spain
3Dipartimento di Scienze Agrarie e Ambientali, Università degli Studi di Milano, 20133, Milano, Italy
Contact the author
Keywords
Grapevine, downy mildew, non-invasive phenotyping tools, imaging, machine vision