Article
Towards Automated Large-Scale 3D Phenotyping of
Vineyards under Field Conditions
Johann Christian Rose
1,
*, Anna Kicherer
2
, Markus Wieland
1
, Lasse Klingbeil
1
,
Reinhard Töpfer
2
and
Heiner Kuhlmann
1
1
Institute of Geodesy and Geoinformation, Department of Geodesy, University of Bonn, Nussallee 17,
53115 Bonn, Germany; wieland@igg.uni-bonn.de (M.W.); klingbeil@igg.uni-bonn.de (L.K.);
heiner.kuhlmann@uni-bonn.de (H.K.)
2
Julius Kühn-Institut, Federal Research Centre of Cultivated Plants, Institute for Grapevine Breeding
Geilweilerhof, 76833 Siebeldingen, Germany; anna.kicherer@julius-kuehn.de (A.K.);
reinhard.toepfer@julius-kuehn.de (R.T.)
* Correspondence: rose@igg.uni-bonn.de; Tel.: +49-228-733-571
Academic Editor: Gonzalo Pajares Martinsanz
Received: 31 October 2016; Accepted: 8 December 2016; Published: 15 December 2016
Abstract:
In viticulture, phenotypic data are traditionally collected directly in the field via visual and
manual means by an experienced person. This approach is time consuming, subjective and prone
to human errors. In recent years, research therefore has focused strongly on developing automated
and non-invasive sensor-based methods to increase data acquisition speed, enhance measurement
accuracy and objectivity and to reduce labor costs. While many 2D methods based on image
processing have been proposed for field phenotyping, only a few 3D solutions are found in the
literature. A track-driven vehicle consisting of a camera system, a real-time-kinematic GPS system for
positioning, as well as hardware for vehicle control, image storage and acquisition is used to visually
capture a whole vine row canopy with georeferenced RGB images. In the first post-processing step,
these images were used within a multi-view-stereo software to reconstruct a textured 3D point cloud
of the whole grapevine row. A classification algorithm is then used in the second step to automatically
classify the raw point cloud data into the semantic plant components, grape bunches and canopy.
In the third step, phenotypic data for the semantic objects is gathered using the classification results
obtaining the quantity of grape bunches, berries and the berry diameter.
Keywords:
viticulture; field phenotyping; 3D point cloud; multi-view-stereo; classification;
berry diameter; number of berries; number of grape bunches
1. Introduction
Grapevine is a perennial crop, and therefore, phenotypic evaluations of yield traits need to
be done directly in the field. Vineyards typically encompass large areas that contain thousands of
single grapevines, with each of these grapevines being able to possess a slightly different phenotype.
In grapevine breeding, the screening of large sets of substantially different genotypes is a special
requirement. Large sets of breeding material are to be screened in their entirety to gather phenotypic
data to be used in the breeding program. Traditionally, phenotypic evaluations on the plant organ
level are either done by visual estimations or destructive sampling. Both methods are time consuming,
often subjective and need to be done by experienced employees. Due to these reasons, the number
of samples is often limited. The development of efficient and objective phenotyping techniques and
high-throughput field phenotyping platforms is crucial to overcome the phenotypic bottleneck [1].
Yield is one of the most commonly-measured and most complex phenotypic traits in viticulture [
2
].
It is defined as the crop weight per vine whereby the weight is dependent on the variation of the
Sensors 2016, 16, 2136; doi:10.3390/s16122136 www.mdpi.com/journal/sensors
