Predicting farrowing of sows housed in crates and pens using accelerometers and CUSUM charts

Tutkimustuotos: ArtikkelijulkaisuArtikkeliTieteellinenvertaisarvioitu

Kuvaus

Piglet mortality is a continuing problem in modern pig production. It is high especially during the first few days after birth and can be reduced by supervision of farrowing.

There is currently a trend to move towards housing without crates due to animal welfare concerns. Thus, automatic systems that can predict farrowing equally well in different housing systems would have wider applicability than ones developed for crated systems only. A method that works with same parameters in different housing systems would be easier to maintain and reuse in different conditions. The aim of this study was to develop a new accelerometer based system to measure sow activity before farrowing. This data was then used to build a model to predict the approach of farrowing, based on increased activity, within 24 h before the start of farrowing.

We used a wireless 3D accelerometer to measure the activity of sows in order to detect farrowing. The accelerometer collars were attached to neck collars of 29 sows farrowing in crates and 33 sows farrowing in pens. We expected sows in pens to show higher activity, but a similar rise in the level.

We used a single model to detect farrowing in both crates and pens. A dynamic linear model was used to model the activity of the sows before farrowing and to extract trend component from seasonal components and a CUSUM chart was used to detect activity increase.

The model detected a rise in activity on average 13 ± 4.8 h (mean ± standard deviation) before farrowing in sows housed in crates and pens with sensitivity of 96.7% and specificity of 100%. The fact that we were able to use a single model with a constant set of parameters gives indication that the method has potential to become a robust indicator of farrowing in different housing systems, even though pre-farrowing activity was higher in sows housed in pens than in crates.
Alkuperäiskielienglanti
LehtiComputers and Electronics in Agriculture
Vuosikerta127
Sivut197-203
Sivumäärä7
ISSN0168-1699
DOI - pysyväislinkit
TilaJulkaistu - syyskuuta 2016
OKM-julkaisutyyppiA1 Alkuperäisartikkeli tieteellisessä aikakauslehdessä, vertaisarvioitu

Tieteenalat

  • 4111 Maataloustiede
  • 413 Eläinlääketiede

Lainaa tätä

@article{19975e2e83014f1eaa3cfcd34f89d42e,
title = "Predicting farrowing of sows housed in crates and pens using accelerometers and CUSUM charts",
abstract = "Piglet mortality is a continuing problem in modern pig production. It is high especially during the first few days after birth and can be reduced by supervision of farrowing.There is currently a trend to move towards housing without crates due to animal welfare concerns. Thus, automatic systems that can predict farrowing equally well in different housing systems would have wider applicability than ones developed for crated systems only. A method that works with same parameters in different housing systems would be easier to maintain and reuse in different conditions. The aim of this study was to develop a new accelerometer based system to measure sow activity before farrowing. This data was then used to build a model to predict the approach of farrowing, based on increased activity, within 24 h before the start of farrowing.We used a wireless 3D accelerometer to measure the activity of sows in order to detect farrowing. The accelerometer collars were attached to neck collars of 29 sows farrowing in crates and 33 sows farrowing in pens. We expected sows in pens to show higher activity, but a similar rise in the level.We used a single model to detect farrowing in both crates and pens. A dynamic linear model was used to model the activity of the sows before farrowing and to extract trend component from seasonal components and a CUSUM chart was used to detect activity increase.The model detected a rise in activity on average 13 ± 4.8 h (mean ± standard deviation) before farrowing in sows housed in crates and pens with sensitivity of 96.7{\%} and specificity of 100{\%}. The fact that we were able to use a single model with a constant set of parameters gives indication that the method has potential to become a robust indicator of farrowing in different housing systems, even though pre-farrowing activity was higher in sows housed in pens than in crates.",
keywords = "4111 Agronomy, 413 Veterinary science, Farrowing , Dynamic linear model , CUSUM chart , Automatic detection , PIGLET MORTALITY , BEHAVIOR , PERFORMANCE , ENVIRONMENT , ONSET , GESTATION , CORTISOL , SENSORS , SYSTEMS",
author = "Matti Pastell and Juha Hietaoja and Jinhyeon Yun and Johannes Tiusanen and Anna Valros",
year = "2016",
month = "9",
doi = "10.1016/j.compag.2016.06.009",
language = "English",
volume = "127",
pages = "197--203",
journal = "Computers and Electronics in Agriculture",
issn = "0168-1699",
publisher = "ELSEVIER SCI IRELAND LTD",

}

TY - JOUR

T1 - Predicting farrowing of sows housed in crates and pens using accelerometers and CUSUM charts

AU - Pastell, Matti

AU - Hietaoja, Juha

AU - Yun, Jinhyeon

AU - Tiusanen, Johannes

AU - Valros, Anna

PY - 2016/9

Y1 - 2016/9

N2 - Piglet mortality is a continuing problem in modern pig production. It is high especially during the first few days after birth and can be reduced by supervision of farrowing.There is currently a trend to move towards housing without crates due to animal welfare concerns. Thus, automatic systems that can predict farrowing equally well in different housing systems would have wider applicability than ones developed for crated systems only. A method that works with same parameters in different housing systems would be easier to maintain and reuse in different conditions. The aim of this study was to develop a new accelerometer based system to measure sow activity before farrowing. This data was then used to build a model to predict the approach of farrowing, based on increased activity, within 24 h before the start of farrowing.We used a wireless 3D accelerometer to measure the activity of sows in order to detect farrowing. The accelerometer collars were attached to neck collars of 29 sows farrowing in crates and 33 sows farrowing in pens. We expected sows in pens to show higher activity, but a similar rise in the level.We used a single model to detect farrowing in both crates and pens. A dynamic linear model was used to model the activity of the sows before farrowing and to extract trend component from seasonal components and a CUSUM chart was used to detect activity increase.The model detected a rise in activity on average 13 ± 4.8 h (mean ± standard deviation) before farrowing in sows housed in crates and pens with sensitivity of 96.7% and specificity of 100%. The fact that we were able to use a single model with a constant set of parameters gives indication that the method has potential to become a robust indicator of farrowing in different housing systems, even though pre-farrowing activity was higher in sows housed in pens than in crates.

AB - Piglet mortality is a continuing problem in modern pig production. It is high especially during the first few days after birth and can be reduced by supervision of farrowing.There is currently a trend to move towards housing without crates due to animal welfare concerns. Thus, automatic systems that can predict farrowing equally well in different housing systems would have wider applicability than ones developed for crated systems only. A method that works with same parameters in different housing systems would be easier to maintain and reuse in different conditions. The aim of this study was to develop a new accelerometer based system to measure sow activity before farrowing. This data was then used to build a model to predict the approach of farrowing, based on increased activity, within 24 h before the start of farrowing.We used a wireless 3D accelerometer to measure the activity of sows in order to detect farrowing. The accelerometer collars were attached to neck collars of 29 sows farrowing in crates and 33 sows farrowing in pens. We expected sows in pens to show higher activity, but a similar rise in the level.We used a single model to detect farrowing in both crates and pens. A dynamic linear model was used to model the activity of the sows before farrowing and to extract trend component from seasonal components and a CUSUM chart was used to detect activity increase.The model detected a rise in activity on average 13 ± 4.8 h (mean ± standard deviation) before farrowing in sows housed in crates and pens with sensitivity of 96.7% and specificity of 100%. The fact that we were able to use a single model with a constant set of parameters gives indication that the method has potential to become a robust indicator of farrowing in different housing systems, even though pre-farrowing activity was higher in sows housed in pens than in crates.

KW - 4111 Agronomy

KW - 413 Veterinary science

KW - Farrowing

KW - Dynamic linear model

KW - CUSUM chart

KW - Automatic detection

KW - PIGLET MORTALITY

KW - BEHAVIOR

KW - PERFORMANCE

KW - ENVIRONMENT

KW - ONSET

KW - GESTATION

KW - CORTISOL

KW - SENSORS

KW - SYSTEMS

U2 - 10.1016/j.compag.2016.06.009

DO - 10.1016/j.compag.2016.06.009

M3 - Article

VL - 127

SP - 197

EP - 203

JO - Computers and Electronics in Agriculture

JF - Computers and Electronics in Agriculture

SN - 0168-1699

ER -