Page 96 - Zbornik_Konf_2019

https://doi.org/10.15414/2019.9788055220703

4 International Scientific Conference Abstracts Book
th
USING OF DRONES FOR CROP PHENOTYPING UNDER BREEDING PROCESS

Taras Kazantsev
Scientific Centre for Aerospace Research of the Earth, National Academy of Sciences of Ukraine, Kyiv,
Ukraine
Drone.UA, Kyiv, Ukraine; E-mail.: antarsih@gmail.com

Modern breeding in agriculture requires testing plenty of phenological traits across
numerous variants and repetitions under field conditions. Experimental fields used for the
plant breeding may contain several thousands of test plots requiring simultaneous
quantitative assessment. Human assessment 'by eye' or traditional laboratory analysis can’t
be effective for this task. Using unmanned aerial vehicles (UAV) also known as drones
equipped with regular and multispectral cameras can be the new effective approach in plant
phenotyping. Images taken from air can be processed for retrieving plenty of plant traits.
Modern drones can be used for aerial survey in fully automatic mode that provides data
uniqueness and allows automatic data processing. However, due to novelty of this approach,
there are no standard procedures of data retrieving and processing yet.
Here we describe the practical experience of using the drones for quantitative
measurement of phenological parameters of plants grown on experimental fields. We aimed
to develop the optimal protocol of aerial survey and methods of data processing for
identification of single plants, estimation number of plants, plant sizes and physiological
status on different growth stages. High accuracy, performance, and reasonable costs were the
main requirements for the developed procedures.
We performed single and regular aerial surveys of rapeseed and wheat crops using
quadrocopters from DJI series equipped with a user-grade RGB camera and Parrot Sequoia
multispectral camera. The experimental fields contained from 3000 to 7000 plots with 1x2 m
size each. The list of tested parameters included: plant count per plot, plant area, plant height,
green/dry ratio, photosynthetic potential, weeds contamination level, and flowering level. We
aimed to optimize survey settings such as flight altitude, flight speed, frequency of image
capturing and camera settings. We also tested the advantages and limitations of the
multispectral camera over a regular RGB camera. Finally, we developed algorithms of data
processing using GIS environment. Drone-derived results were compared with data obtained
by manual in-field tests and manual image inspections.
The results indicate high efficiency of using drones for plant phenotyping and advantage
of this method over traditional in-field methods. Drone-based methods demonstrated about
90 % accuracy in assessing of plant parameters. For metric parameters such as plant count
and plant sizes as well as for flowering level, the regular RGB camera was more effective
comparing to the multispectral camera because of bigger frame size. Recommendations for
the optimal data quality as well as of limitations of the drones are discussed.

Keywords: Drones, UAV, Plant phenotyping, experimental fields, breeding.

|96 4 International Scientific Conference Agrobiodiversity Nutrition, Health and Quality of Human and Bees Life
th
September 11–13, 2019