Modeling and Calibrating Visual Yield Estimates in Vineyards

Abstract

Accurate yield estimates are of great value to vineyard growers to make informed management decisions such as crop thinning, shoot thinning, irrigation and nutrient delivery, preparing for harvest and planning for market. Current methods are labor intensive because they involve destructive hand sampling and are practically too sparse to capture spatial variability in large vineyard blocks. Here we report on an approach to predict vineyard yield automatically and non-destructively using images collected from vehicles driving along vineyard rows. Computer vision algorithms are applied to detect grape berries in images that have been registered together to generate high-resolution estimates. We propose an underlying model relating image measurements to harvest yield and study practical approaches to calibrate the two. We report on results on datasets of several hundred vines collected both early and in the middle of the growing season. We find that it is possible to estimate yield to within 4% using calibration data from prior harvest data and 3% using calibration data from destructive hand samples at the time of imaging.

Vineyard yield estimation by automatic 3D bunch modelling in field conditions

“…Finally a clustering is applied to remove false positives. In a recent work (Nuske et al, 2012) the use of calibration data from prior harvest data enhanced the previous results. Moreover, 2D computer vision techniques have also been applied to individual strains for the identification of plant elements (Herrero Langreo et al, 2010) or counting individual berries using a flatbed scanner (Battany, 2008).…”

“…This volume limited by the hull (Vc) collects half of the whole cluster due to the field of view of the camera and the limitations of the branches and leaves, closing the hidden side of the cluster by a flat surface. Because the convex hull includes empty spaces where there are no berries, as well as the exclusivity of the visible side of the bunch it is necessary to include an empirical correction factor (K) (Nuske et al, 2011;Nuske et al, 2012) that will refine the estimated volume (Ve).…”

Potentials of small, lightweight and low cost Multi-Echo Laser Scanners for detecting Grape Berries

“…As alternative, a method to detect and count grape berries by exploiting shape and visual texture in images has been proposed Nuske et al, 2012). The selection of these visual features directly addresses the crucial issues of different lighting and lack of color contrast.…”

Automatic grape bunch detection in vineyards with an SVM classifier

“…So an automatic yield estimation system based on image processing in vineyards has become a mainstream research topic in recent years. For tackling this problem, Nuske et al [3] developed a prediction method by detecting the berry number from images taken in a vineyard. This paper applied a Radial Symmetry Transform (RST) [4] to find keypoints, which are potential berry locations, from every single image as the initial step for later feature analysis.…”