Prediction of yield by digital image analysis of vine

Abstract

The grape yield per vine of cv. Cabernet Sauvignon (Vitis vinifera L.) was evaluated on the basis of digital image processing of vine part. Digital camera was mounted on tripod and used for taking photos of 1 x 1 m portions of canopy. The Adobe Photoshop software was used to analyse image for the colour counting of the blue pixels of grape in the quadrant region. The actual yield was obtained from the photographed vines by hand harvesting of sampled portions. Linear regression was used for calculation of the correlation between blue pixels and grape weight. The relatively strong relationship between blue pixels and grape weight (R2=0.91) was obtained. Based on these results, we can recommend this simple technique for yield forecasting. [Projekat Ministarstva nauke Republike Srbije, br. TP31063]

Keywords

• grape weight
• colour analysis
• number of pixels