Grapevine bunch volume estimation using an infrared depth sensor under laboratory and field conditions

Abstract

An accurate yield estimation is a constant demand in the winemaking industry. Current methods are destructive and time-consuming. In general, yield prediction is performed using weather and historical yield data combined with manual samplings in the field. The aim of this study is to define a novel method to estimate bunch volume with a non-destructive technique. In this study 3D images of 21 grapevine bunches (Vitis vinifera L. cv. Shiraz) were obtained by a Kinect infrared depth sensor using three different approaches: i) laboratory condition (standard light environment); ii) field condition full canopy, prior to leaf removal and iii) field condition after leaf removal (100% of leaf removal in the bunch zone). The bunch volume obtained by the traditional water displacement method was used as a reference value. Results show a high correlation between the reference volume and the estimated volume obtained by 3D images in laboratory condition with a coefficient of determination (r²) of 0.93. In field conditions, the leaf removal treatment was the best result for estimating the bunch volume with an r² value of 0.82. Therefore, it is possible to ensure that the use of an infrared depth sensor is a valid method to estimate bunch volume in both laboratory and field condition. This novel approach could be the first step in a vineyard yield estimation program if an elaborate sampling method is provided.