Presentation delivered at European Conference for Precision Livestock Farming at 11 September 2013 by Andrew Spink. How GPS and 3D accelerometers help in sensing and analysing different animal behaviours.

1. Animal behaviour analysis with GPS
and 3D accelerometers
6th European Conference on Precision Livestock
Farming
Andrew Spink, Brian Cresswell, Andrea Kölzsch,
Frank van Langevelde, Marjolein Neefjes,
Lucas Noldus, Herman van Oeveren, Herbert
Prins, Tamme van der Wal, Nelleke de Weerd &
Willem Frederik de Boer

2. E-Track Consortium
• Noldus Information Technology BV
• AeroVision BV
• Biotrack Ltd
• Wageningen University
• Leibniz Institute for Zoo and Wildlife Research
• Netherlands Institute of Ecology
11-Sep-13 2

3. Better and more reliably track and analyse the
movements and behaviour of animals under
field conditions
• Limited battery life means sample rate has to be
low — Smaller and more efficient GPS receivers
• Inefficiencies of data import, formatting,
smoothing, etc — Complete integrated solution of
software and hardware
• Location is the start, knowing what the animal is
doing is more relevant —EGNOS-enhanced GPS
to detect animal behaviours
11-Sep-13 3

4. 11-Sep-13 4
Position of cow determined by GNSS (GPS)
Ionosphere causes inaccuracies
EGNOS ground station
measures difference in
actual position and GNSS
position
Correction factor transmitted live
by Galileo satellite and offline by
internet (EDAS)

5. • European Geo-stationary
Navigation Overlay Service
• Accuracy improves from
about 10m to 1-2m
• Factors such a tree canopies
and deep valleys will still
cause inaccuracies (less
satellites)
• More accurate tracking
opens the possibility to go
beyond simple location data
to detection of behaviours
11-Sep-13 5

6. Livestock Experiment
Methods
• Cows equipped with GPS collars and manually
observed with Pocket Observer
• GPS data visualized with TrackLab
• Correlation between behaviours & GPS data
tested with ANOVA
• Decision tree used to
create a model, which
was then validated on
new data
11-Sep-13 6

7. Livestock Experiment
• Cow moves slower near water and silage
• Searching behaviour (long segments, high speed)
cf foraging behaviour (short segments, low
speed)11-Sep-13 8
Results

8. Livestock Experiment
Results (raw data)
• Ruminating and
Standing are not
different
• Foraging, Walking,
do differ from
Ruminating/
Standing
11-Sep-13 9
Turnangle°

9. Livestock Experiment
Results (raw data)
• Walking has higher velocity than other behaviours
• Foraging & Walking have lower turn angle than
Ruminating and Standing
11-Sep-13 10
Turnangle°
Distance(m)
Foraging
Ruminating
Walking
Standing

10. Livestock experiment
0
10
20
30
40
50
60
70
80
90
100
Foraging Ruminating Standing Walking Standing &
Ruminating
11-Sep-13 13
% correctly classified

11. Livestock Experiment
Results – Effects of
forest canopy
• Spread of data is
less than in open
area
• Groups still different
• Correct classification
is only 57%
11-Sep-13 14

12. GPS & Accelerometer data
Methods
• EGNOS GPS collars attached to
geese (<2% body weight)
• 3D accelerometers on collars
(not just tilt switch)
• Behaviours manually scored
with The Observer XT
• Preliminary data
– visual comparisons
11-Sep-13 15

13. GPS & Accelerometer data
11-Sep-13 16

14. GPS & Accelerometer data
11-Sep-13 17
Behaviours which cannot be
distinguished by GPS, have
different accelerometer
waveforms

15. Ongoing research
• TrackLab will be expanded to include import of
accelerometer data, data segmentation, analysis
of social behaviours
• We are very interested in feedback!
• Combined GPS, accelerometer & other data for
detection of behaviours
• Library of behaviours & parameters to detect
them
11-Sep-13 18

16. Ongoing research
• Intelligent geofencing — accurate positions,
behaviours as input
• Combination of indoor and outdoor tracking
• Partnership with researchers working with
precision livestock
11-Sep-13 19

17. GPS & Accelerometer data
Conclusions
• GPS data can be used to distinguish a variety of
behaviours
• The more accurate the GPS, the more behaviours
• Additional data such as from accelerometers can
enable detection of more behaviours
— more research needed
• TrackLab is an efficient
solution for acquiring,
visualizing and analysing
tracking data
11-Sep-13 20

18. E-Track is carried out in the context of the Galileo FP7 R&D programme
supervised by the GSA. (nr. 277679-2). For more information, please check
www.etrack-project.eu .
Thank you for your attention.
11-Sep-13 21