Content uploaded by Anita Kapun
Author content
All content in this area was uploaded by Anita Kapun on Jun 22, 2016
Content may be subject to copyright.
CIGR-AgEng conference Jun. 26–29, 2016, Aarhus, Denmark
∙ 1 ∙
Test of a UHF-RFID system for health monitoring of finishing pigs
Anita Kapun a,*, Felix Adrion a, Laura Alena Schmid a, Max Staiger a, Eva-Maria Holland a, Eva Gallmann a,
Thomas Jungbluth a
a Institute of Agricultural Engineering, University of Hohenheim, 70599 Stuttgart, Germany
* Corresponding author. Email: anita.kapun@uni-hohenheim.de
Abstract
Due to increasing herd sizes, the demand for automatic health and activity monitoring systems for individual animals
is getting stronger. The aim of this study is to fulfill this task by using a validated ultra-high frequency radio frequency
identification system (UHF-RFID) to monitor the visiting events (time, frequency, duration) of finishing pigs at so-called
hotspots in a pen. This contribution presents initial analyses as a step toward this goal.
Two groups of finishing pigs (33 in total) were tagged with UHF-RFID transponder ear tags and were kept in two
pens of an outdoor climate stable equipped with UHF readers at the trough, the drinker, the playing material and the door
to a yard area. Over a period of nearly 12 weeks, the visiting events of each pig at these locations were monitored via a
monitoring software. Additionally, their health status (lameness, tail lesions, skin lesions, diarrhea, coughing and
shortness of breath) was observed twice a week.
The average daily duration of visits at the trough across all 33 pigs was 145 min. The pig with the shortest duration
was, on average, 101.3 min per day at the trough, the one with the longest duration 203.7 min. Though there was no
difference in the average feeding duration of healthy pigs and pigs classified as sick in general, a finer classification
showed that pigs that were declared as lame (123.9 min) or coughing (113.3 min) had shorter daily visiting time at the
trough. When looking at the single RFID events, all pigs showed circadian rhythms of activity on every hotspot. Lame
pigs displayed a longer night-time break on days with lameness than on healthy days. Also, the influence of the fattening
day on the duration at the trough was regarded.
The results promise a possibility to detect lameness based on monitored changes of behavior patterns. Further
analyses are planned, e.g. on other hotspots and pigs with observed diarrhea with the objective of forecasting diseases.
Keywords: health, behavior, monitoring, fattening pigs, ultra-high frequency, radio frequency identification
1. Introduction
Due to increasing herd size per production unit health monitoring is becoming a growing challenge in modern pig
husbandry. Hence, systems for automatic animal health and activity monitoring would be of great help to the farmer.
Observing anomalous behavior could be a promising method for the detection of health impacts. Pigs’ visits at different
hotspots in the pen (e.g. at the trough) can be registered by using RFID (Radio Frequency Identification) for electronic
animal identification (Adrion et al., 2015; Maselyne et al., 2014a; Maselyne et al., 2015). Analyzing a pig’s duration and
frequency of visits at a hotspot may lead to conclusions on the animal’s health. The use of UHF technology (ultra-high
frequency) instead of LF (low frequency) enables a simultaneous identification without separation and a larger reading
range (Adrion et al., 2015; Barge et al., 2013). The aim of this study was to develop an automatic health and activity
monitoring system based on UHF-RFID and the visiting events at the trough, the drinker, the door to the yard area and
the playing material. This contribution shows initial analyses in order to examine the suitability of different indicator
variables and covariates with influence on the pigs’ behavior for the modelling that serves as basis for a monitoring
system.
2. Materials and Methods
The study was carried out in Germany in an outdoor climate stable with insulated lying boxes and a yard area (“Pig-
Port 3”) for pigs. Two mixed-gender groups of finishing pigs (33 in total) were tagged with UHF-RFID transponder ear
tags and kept in two similarly structured pens (Figure 1). The pigs (German Landrace, DanBred, Saddlebag pigs)
weighed on average 55 kg at the beginning of the tests and were fattened to a final weight of about 120 kg. Each pen was
divided into different functional areas, namely into the resting area with solid surface and minimal litter, the fully slatted
feeding and drinking area with a wet feeder and three nipple drinkers and the fully slatted yard area where the pigs
usually defecate. A metal chain with a piece of wood and a plastic tube was located in the yard area and used as playing
material. The hotspots feeding area (trough), drinking area (drinker), door to the yard area (door) and playing area (toy)
in both pens were equipped with UHF antennas (high gain patch antennas at the trough, drinker and the toy, cable
antennas in plastic tubes at the door). The readers used in this UHF-RFID system were functional models (deister
electronic GmbH, agrident GmbH, Barsinghausen, Germany) with a multiplexer for 4 antennas, a maximum output
power of 29 dBm and an operation frequency of 865.7 MHz. The UHF ear tags were developed within the research
project especially for the use with pigs. The transponders were equipped with an Impinj Monza 4® chip and had a PIF
CIGR-AgEng conference Jun. 26–29, 2016, Aarhus, Denmark
∙ 2 ∙
antenna design (Planar Inverted F-Shaped Antenna; Adrion et al.; 2015). They were sized approx. 30 x 40 mm and
grouted into a flexible plastic ear tag (Primaflex®, Caisley International GmbH, Bocholt, Germany).
Figure 1. Floor plan of the two test pens for the study.
Preliminary tests were carried out to figure out the most suitable antenna power for each hotspot to detect an animal’s
visit correctly. For each hotspot four different settings of the antenna power (23-28 dBm) were tested by validating the
RFID events with video recordings of 8 focal pigs. It should be clarified, that the video analysis was based on the pigs’
location in the reading range and not on any interaction with the hotspots such as feeding at the trough. The aggregation
of RFID events – i.e. by which criteria single readings are summed up to events – was optimized for each hotspot and
each tested antenna power. These visiting events were compared to the video recordings (at least two hours of high
activity for each hotspot and each tested antenna power) with statistical measures such as true positive rate (sensitivity)
and true negative rate (specificity). For example, a positive event (video event) was classified as true, if the pig in
question was registered at the hotspot some time during this event (RFID event). Based on these results, the antenna
powers and RFID event aggregation parameters for each hotspot were selected as follows for the main study (Table 1).
Sensitivity was very good at the trough and the toy, however, at the drinker and the door potential for further
optimization of the antenna field was revealed.
Table 1. Antenna power and RFID event aggregation for every hotspot in the main study.
Hotspot
Antenna
power
[dBm]
RFID event aggregation
Statistical measures*
Maximum gap between
two single readings
[s]
Minimum
duration of event
[s]
Sensitivity
[%]
Specificity
[%]
Precision
[%]
Accuracy
[%]
Trough
24
50
1
92.0
97.5
97.2
94.8
Drinker
24
60
1
79.0
92.0
90.1
85.7
Toy
24
30
1
100.0
97.8
97.4
98.8
Door
26
30
0
48.5
93.0
86.9
71.1
*Sensitivity = true positive rate, Specificity = true negative rate, Precision = positive predictive value
The main study ran for about 12 weeks. During this period the UHF-RFID system consisting of UHF ear tags, UHF
antennas, readers and a monitoring software (Phenobyte GmbH, Ludwigsburg, Germany) that recorded and aggregated
occurring RFID events ran constantly (except on one day due to a broken USB interface). Also temperature and humidity
inside and outside the building were permanently logged by dataloggers (testo 175H, Testo AG Lenzkirch, Germany).
Furthermore, the health status of the pigs was observed twice a week in addition to the usual daily control by the farm
CIGR-AgEng conference Jun. 26–29, 2016, Aarhus, Denmark
∙ 3 ∙
manager. The rating of the health status included lameness, tail lesions, skin lesions, diarrhea, coughing and shortness of
breath – each divided into 2 to 4 grades of severity. Apart from this, the management in the test pens was the same as in
the other pens on this farm, so that the test was carried out under conditions of good agricultural practice.
Only complete trial days were used for the analysis. That means, each day a pig could be monitored completely by the
UHF-RFID system, beginning with the day after the trial started until the day before the pig was removed from the stable
for slaughter, was considered for the calculations, apart from days with incomplete data because of technical reasons. The
following analyses were performed to assess the possibility of an early detection of diseases on the basis of RFID events
by combining these events with the health conditions. Basis of these analyses were the aggregated RFID events at every
hotspot and the resulting daily mean values of the duration and frequency of visits. Based on the ratings of health status,
the pigs were classified into two groups. Pigs that showed any lameness or diarrhea or a higher grade of coughing or skin
lesions were classified as sick. Other pigs were declared as healthy.
3. Results and Discussion
The daily mean values of the duration and frequency of visits at the different hotspots are summarized in Table 2. On
average, the frequency at the trough was 35.0 visits per day with an average duration of 145.0 minutes. The average
visiting events at the drinker were 26.6 times per day with a duration of 24.7 minutes. The toy was the least frequented
hotspot with on average 9.9 visits per day (8.2 min d-1). The door was the hotspot with the shortest average duration
(5.7 min d-1), but was visited relatively often (23.3 visits d-1). The minima and maxima are obtained from every single
daily mean value of every pig. However, regarding the minima and maxima of daily mean values grouped by pigs, a
strong inter-animal variability is reflected. For example, the pig with the shortest average daily duration at the trough was
101.3 min per day at this hotspot, the one with the longest daily duration was there on average 203.7 min. The variability
between the pigs was particularly large at the drinker with an average daily visiting time between 8.0 min and 56.0 min.
Maselyne et al. (2014b) also found variations within and between finishing pigs for the daily duration of different
behaviors. The average daily durations of feeding (99.4 min) and drinking (15.8 min) were shorter than in this study here,
but were only observed during the day between 7 AM and 9 PM.
Table 2. Daily mean values (MV) of duration and frequency of visits (n = 2365, pigs×complete trial day).
Hotspot
Parameter
MV
Min.
Max.
Trough
Duration [min d-1]
145.0
22.0
344.8
Frequency [1 d-1]
35.0
7
103
Drinker
Duration [min d-1]
24.7
0.2
174.7
Frequency [1 d-1]
26.6
1
82
Toy
Duration [min d-1]
8.2
0
173.0
Frequency [1 d-1]
9.9
0
62
Door*
Duration [min d-1]
5.7
0.01
133.3
Frequency [1 d-1]
23.3
1
106
*n = 2205 due to technical defect
In the present study, the pigs were classified into two groups based on the observed health status: sick and healthy.
Figure 2 shows the average daily visiting time at the trough based on different health conditions (only health observation
days were considered for this figure). The daily mean value of pigs classified as sick was 145.2 min, 141.5 min of pigs
classified as healthy. As the figure implies, there is a clear difference between the average daily duration at the trough of
pigs that were lame (123.9 min) or coughing (113.3 min) and healthy pigs. For example, pig 43 had the most lameness
days with 4 out of 24 health observation days. The average duration and number of visits of this pig at the trough per day
were both higher on days were no lameness was detected (104 compared to 87 min and 32 compared to 23 visits). The
fact that the drinking or feeding behavior of pigs can be affected by lameness or other diseases is supported by the
findings in other studies (Madec et al., 1986; Junge, 2015; Weary et al., 2009). Pigs with diarrhea and skin lesions had an
average daily feeding time of 147.8 min respectively 145.6 min. Consequently, they hardly showed any difference to the
mean value of healthy pigs (141.5 min). Though the results seem to suggest that lame and coughing pigs eat less, they are
not necessarily indicative due to the small quantity of occurrences of diseases. Nevertheless, the results are very
promising. They allow the assumption, that lameness and cough could be detected by an RFID system based on the
visiting events at the trough. Further studies with an UHF-RFID monitoring system are needed to confirm this
assumption.
CIGR-AgEng conference Jun. 26–29, 2016, Aarhus, Denmark
∙ 4 ∙
Figure 2. Average daily visiting time at the trough based on different health conditions.
When looking at the single visiting events instead of the mean values per day, the data indicates circadian rhythms for
the visits of every hotspot, which is also supported by other studies. For example, Ingram and Dauncey (1985) found that
the peak of food intake was significantly greater in the light than in the dark and water intake was correlated very
strongly with that. Overall, the activity of pigs in the present study is greater during the day and there is often a night-
time break of several hours where there is no visiting event at any hotspot. Despite that, a night time break that is longer
than usual could be an indicator for diseases, because sick animals may tend to be sleepy or inactive and less motivated to
move (Hart, 1988). For example, in 6 out of 10 times where lameness was detected in the present study, the night time
break at the trough was found to be more than 8 h long on the day of health observation or ±1 day. This was the case on
only 14 % of the days where no lameness was observed at these pigs. This circumstance should be further investigated by
studies with a larger number of animals and with minimized potential sources of error by eliminating time gaps in health
observation.
With regard to possible interrelations between health condition and frequency or duration of visits at a hotspot, other
influencing variables must not be forgotten, such as atmospheric humidity and temperature or the age and thus the weight
of the pigs. As an example, figure 3 shows the decreasing trend over time of the duration of visits at the trough of every
pig (daily mean values) in relation to the time of fattening. In contrast to that, Brown-Brandl et al. (2013) observed
increasing time at the feeders until day 95-105 of age with a maximum of only 76.7 min d-1. In general, a possible
influence on the results by aspects of the social behavior and interaction (social facilitation) must not be forgotten, but
could not be investigated in this study yet. For example, if a pig fed ad libitum sees another pig eating, it will be
motivated to eat again (Deen, 2010).
CIGR-AgEng conference Jun. 26–29, 2016, Aarhus, Denmark
∙ 5 ∙
Figure 3. Daily mean value of duration of visits at the trough of every pig and trend over fattening time.
4. Conclusions
The present study shows much potential in finding sick finishing pigs based on UHF-RFID detection. The RFID
events display the behavior of the pigs represented by the daily duration and frequency of stay at the trough, the drinker,
the toy and the door. The analysis of the data and correlations so far indicate that lameness and coughing could be
detected automatically through this method. The analysis of the single RFID events and circadian rhythms seem to be
very promising to find suitable variables for disease detection. Further studies and additional analyses of the data are
planned at other hotspots and involving influencing variables such as atmospheric humidity and temperature. It will be
investigated, for example, whether other diseases can be detected by regarding drinking events or how visiting events on
different hotspots are correlated to each other. The influence of social facilitation is another point of interest. The aim is
to develop a prediction model based on a UHF-RFID system which could be used for early detection of diseases in
finishing pigs.
Acknowledgements
The project was supported by funds of the Federal Ministry of Food and Agriculture (BMEL) based on a decision of
the Parliament of the Federal Republic of Germany via the Federal Office for Agriculture and Food (BLE) under the
innovation support programme. FKZ 28154T0910.
References
Adrion, F., N. Hammer, F. Eckert, S. Götz, and E. Gallmann, 2015. Adjustment of a UHF-RFID system for hotspot
monitoring of fattening pigs. In Precision Livestock Farming ’15. Papers presented at the 7th European Conference on
Precision Livestock Farming. Milan, Italy, September 15-18: Eds., M. Guarino and D. Berckmans. 573-582.
Barge, P., P. Gay, V. Merlino, C. Tortia, 2013. Radio frequency identification technologies for livestock management
and meat supply chain traceability. Canadian Journal of Animal Science 93 (1), 23-33.
http://dx.doi.org/10.4141/cjas2012-029.
Brown-Brandl, T.M., G.A. Rohrer, R.A. Eigenberg, 2013. Analysis of feeding behavior of group housed growing–
finishing pigs. Computers and Electronics in Agriculture 96, 246–252. http://dx.doi.org/10.1016/j.compag.2013.06.002.
Deen, J., 2010. Pigs: Behavior and Welfare Assessment. In Encyclopedia of animal behavior. Oxford: Elsevier
Science, Ed., M.D. Breed. 731–739.
Hart, B.L., 1988. Biological basis of the behavior of sick animals. Neuroscience & Biobehavioral Reviews 12 (2),
123–137. http://dx.doi.org/10.1016/S0149-7634(88)80004-6.
Ingram, D.L., M.J. Dauncey, 1985. Circadian rhythms in the pig. Comparative Biochemistry and Physiology Part A:
Physiology 82 (1), 1–5. http://dx.doi.org/10.1016/0300-9629(85)90695-4.
Junge, M., 2015. Verhaltens- und Gesundheitsmonitoring für die Gruppenhaltung tragender Sauen. Dissertation,
University of Hohenheim, Stuttgart, Germany.
CIGR-AgEng conference Jun. 26–29, 2016, Aarhus, Denmark
∙ 6 ∙
Madec, F., R. Cariolet, R. Dantzer, 1986. Relevance of some behavioural criteria concerning the sow (motor activity
and water intake) in intensive pig farming and veterinary practice. Annales de Recherches Vétérinaires 17 (2), 177-184.
Maselyne, J., W. Saeys, B. De Ketelaere, K. Mertens, J. Vangeyte, E.F. Hessel, S. Millet, A. Van Nuffel, 2014a.
Validation of a High Frequency Radio Frequency Identification (HF RFID) system for registering feeding patterns of
growing-finishing pigs. Computers and Electronics in Agriculture 102, 10-18.
http://dx.doi.org/10.1016/j.compag.2013.12.015.
Maselyne, J., W. Saeys, B. De Ketelaere, P. Briene, S. Millet, F. Tuyttens, A. Van Nuffel, 2014b. How do fattening
pigs spend their day? In Proceedings of the 6th International Conference on the Assessment of Animal Welfare at the
Farm and Group Level. Clermont-ferrand, France, September 3-5: Eds., L. Mounier and I. Veissier.
http://dx.doi.org/10.3920/978-90-8686-798-1.
Maselyne, J., I. Adriaens, T. Huybrechts, B. De Ketelaere, S. Millet, J. Vangeyte, A. Van Nuffel, W. Saeys, 2015.
Measuring the drinking behaviour of individual pigs housed in group using radio frequency identification (RFID).
Animal, May 2015.
Weary, D.M., J.M. Huzzey, M.A.G. von Keyserlingk, 2009. Board-invited review: Using behavior to predict and
identify ill health in animals. Journal of animal science 87 (2), 770–777. http://dx.doi.org/10.2527/jas.2008-1297.