Phenoliner 2.0: RGB and near-infrared (NIR)image acquisition for an efficient phenotypingin grapevine research

In grapevine research, phenotyping needs to be done
for different traits such as abiotic and biotic stress. This phenotypic
data acquisition is very time-consuming and subjective
due to the limitation of manual visual estimation. Sensor-based
approaches showed an improvement in objectivity and throughput
in the past. For example, the ‘Phenoliner’ a phenotyping
platform, based on a modified grape harvester, is equipped with
two different sensor systems to acquire images in the field. It
has so far been used in grapevine research for different research
questions to test and apply different sensor systems. However,
the driving speed for data acquisition has been limited to 0.5
- 1 km/h due to capacity of image acquisition frequency and
storage. Therefore, a faster automatic data acquisition with high
objectivity and precision is desirable to increase the phenotyping
efficiency. To this aim, in the present study a prism-based simultaneous
multispectral camera system was installed in the tunnel
of the ‘Phenoliner’ with an artificial broadband light source for
image acquisition. It consists of a visible color channel from 400
to 670 nm, a near infrared (NIR) channel from 700 to 800 nm, and
a second NIR channel from 820 to 1,000 nm. Compared to the
existing camera setup, image recording could be improved to at
least 10 images per second and a driving speed of up to 6 km/h.
Each image is geo-referenced using a real-time-kinematic (RTK)-
GPS system. The setup of the sensor system was tested on seven
varieties (Riesling, Pinot Noir, Chardonnay, Dornfelder, Dapako,
Pinot Gris, and Phoenix) with and without symptoms of biotic
stress in the vineyards of Geilweilerhof, Germany. Image analysis
aims to segment images into four categories: trunk, cane, leaf,
and fruit cluster to further detect the biotic stress status in these
categories. Therefore, images have been annotated accordingly
and first results will be shown.