At TUS Nordics 2017, Philipp Trénel gave the presentation ‘Drones-as-a-Service for agricultural applications’ in our Arctic track, on Thursday 12 October 2017.
Drones-as-a-Service for agricultural applications (by Philipp Trénel)
1. Philipp Trénel, Søren K. Boldsen, Thomas Nitschke | DTI - Danish Technological Institute, AgroTech
Drones-as-a-Service for agricultural applications
experiences and findings from drone flights aimed at detecting plant diseases
2. the setting
the use of civil drones is expected to increase drastically over the next years(1) with an expected
150.000 new jobs and a turnover of 15 billion Euros reached before year 2050 in Europe alone(2,3)
agriculture is often highlighted as one of the business areas with a great business potential for
drones(2,4)
drones are expected to become an integrated part of future precision agriculture (PA)
techniques(5)
Drones-as-a-Service for agricultural applications
1) Global drone teknologi. Rapport af Teknologisk Institut udarbejdet for Styrelsen for Forskning og Innovation som led i RK kontrakt om professionelle civile droner, 2016.
(2) Droner – en ny vækstbranche? En analyse af jobpotentialet ved ubemandede fly i Danmark. Oxford Research, 2015.
(3) Fremtidens regulering af civile droner, Rapport fra en tværministeriel arbejdsgruppe, Trafikstyrelsen, 2015
(4) Civile droner i Danmark – potentialer, udfordringer og anbefalinger. Rapport af Teknologi Rådet udarbejdet for Uddannelses- og Forskningsministeriet, 2014.
(5) Zhang, C. & Kovacs, J.M. 2012. The application of small unmanned arial systems for precision agriculture: a review. Precision Agric 13:693-712.
3. precision agriculture (PA)
is expected to be part of the solution for present and future food, agriculture, climate and
environmental challenges(6,7)
aims at automated & area-specific optimization of agricultural operations based on small-
scale sensor measurements, while at the same time minimizing the environmental costs
drones may provide a time-effective technical solution for an automatic sampling of dense sensor
data across large areas
but, only few Danish farmers are actively using drones or drone-based companies and their
products for improving their operations in the field(8)
Drones-as-a-Service for agricultural applications
(6) EIP-AGRI Focus Group Precision Farming – EU final report 2015
(7) Precision agriculture: an opportunity for EU farmers. EU report 2014.
(8) Kortlægning af droner i Danmark. Rapport af Teknologisk Institut, 2016.
4. why are drones not ubiquitous in PA applications, yet?
restrictive laws, not permitting full automation and BVLoS*
high degree of technology maturation needed
high degree of user-friendliness needed(9)
value of drone-assisted technologies in commercially driven farms still not demonstrated and
validated in a Danish setting
drones must act as an integral part of a full-solution service product, i.e. drones-as-a-service
(DaaS)(8)
Drones-as-a-Service for agricultural applications
*: BVLoS = Beyond Visual Line of Sight
(8) Kortlægning af droner i Danmark. Rapport af Teknologisk Institut, 2016.
(9) Fountas, S., Pedersen, S.M. and Blackmore, S. 2005. ICT in Precision Agriculture – Diffusion of
Technology. In: E.Gelb and A. Offer (Eds.), ICT in Agriculture
5. DaaS for agricultural applications
requires an intelligent integration of
drone, computer vision, machine learning
algorithms and user-friendly web platforms
with agronomic domain knowledge
requires large amounts of ground-truth
data to be sampled, before any valid and
robust prediction algorithm for a PA
solution can be build
multi-domain nature!
Drones-as-a-Service for agricultural applications
8. Drones-as-a-Service for agricultural applications
DaaS for agricultural applications
DaaS IT-platform
2. Data acquisition
1. User request
3. Data 4. Stitching
2. Time and flight plan, sensor choice
Data acquisition
Ground truth data acquisition
Stitching
Automated trial plot detection
Library: {ground truth, sensor data}
Training+testing predictive models
Calibration
Prediction model
5. User-oriented
presentation
6. Automation/action
External data
field 100-1
9. The Danish Technological Institute (DTI)
is actively contributing to the promotion
and innovation of civil drones in the
Danish context by:
- providing a test-center at Odense
Robotics
- implementing shared autonomy into
drone technologies
- producing 3D-prints for drone
components
- developing alternative energy
systems for prolonged flying time
- contributing to the ministerial working
group application of drones in
agriculture for the Ministry of
Environment and Food of Denmark
Danish Technological Institute (DTI)
the Danish Technological Institute is part of the solution!
brings together experts from the fields of robotics, agriculture,
computer science and machine learning.
conducting 1000+ field trials/yr. in collaboration with SEGES
10. on-going activities
two DaaS projects, aiming at detecting and quantifying plant diseases in two different crops
(wheat and sugar beet) based on data from drone-borne multispectral sensors and ground-truth
data collected in two Danish field trials
demonstrating the complex multi-domain nature of DaaS applications in the agricultural setting,
their challenges and solutions.
11. DTI: Danish Technological Institute
Wheat variety trial Sugar beet trial
Ground truth
Trial locality Koldkærgård, eastern Jutland Gedsergård, Falster, Southern Denmark
Trial owner
SEGES, Landsforsøg® Nordic Beet Research
Measures Yield (hkg/ha)
Protein content at harvest (% of DM)
Starch content at harvest (% of DM)
Septoria (%)
Mildew (%)
Yellow rust (%)
Mildew (%)
Ramularia (%)
Beet rust (%)
Cercospora (%)
Drone Ebee Phantom 3 DJI S1000+ octo-rotor
Operator Integra Bo JM Secher, Nordic Sugar DTI
Sensor Multispectral Multispectral Hyperspectral
Brand Parrot Sequoia Parrot Sequoia Specim FX10
Spectral bands
(nm)
Green, Red, Red edge, NIR
510-590, 620-700, 725-745, 750-830
Green, Red, Red edge, NIR
510-590, 620-700, 725-745, 750-830
224 spectra
397 - 1005
Data management
Stitching Pix4D mapper Pix4D mapper Pix4D mapper
Plot detection R-function, DTI R-function, DTI R-function, DTI
Data analysis R, DTI R, DTI R, DTI
14. Drones-as-a-Service for agricultural applications
challenges
holding the time plan
stitching is challenged when drone data are collected at sub-optimal weather conditions
trial plot detection: retrieving pixel values for each trial plot requires an automated algorithm
dimension reduction of multi-/hyperspectral data, Nground truth ≪ Ndrone data: the curse of
dimensionality
honest model evaluation using valid hold-out data, simulating prediction for unknown fields,
varieties, years, etc.
15. Drones-as-a-Service for agricultural applications
conclusions
buidling DaaS solutions for agricultural applications is a vastly complex task
its multi-domain nature does not offer simple ‘low-hanging fruits’ solutions
DTI is currently analyzing data from two DaaS projects aiming at detecting plant diseases
17. This activity is funded by the Danish Ministry of Higher Education and Science as part of the project
“Drones – from development to implementation”, for more information visit www.teknologisk.dk
Thank you!