Background:Grasslands are the primary source of feed for grazing livestock,and as such,knowledge on how to best manage livestock and grasslands,through the use of spatiotemporal modelling,will assist in the long-term ...Background:Grasslands are the primary source of feed for grazing livestock,and as such,knowledge on how to best manage livestock and grasslands,through the use of spatiotemporal modelling,will assist in the long-term management of a valuable ecosystem resource.Methods:This study was conducted over 14 months between March and April 2017 in Orange,NSW,Australia.The study evaluated sheep behaviour in relation to the presence of pasture species,environment and paddock structures,using random forest modelling,to predict sheep location under continuous high(HSR,13 DSE ha−1)and low(LSR,7DSE ha−1)stocking rates.Results:In the LSR,significant drivers included water,shade and fence lines(p<0.01).In the HSR,only fence lines and available biomass were found to be significant(p<0.01).The presence of green legumes in both stocking rates often increased residency by sheep.Animals spent more time together in the LSR,suggesting that social behaviour played a larger role than pasture quantity and quality in driving grazing behaviours.Conclusions:Understanding how pasture type can influence grazing behaviours and also how animal behaviour affects pasture performance and utilisation is important in developing long-term sustainable management strategies on a paddock scale.展开更多
Classification of sheep behaviour from a sequence of tri-axial accelerometer data has the potential to enhance sheep management.Sheep behaviour is inherently imbalanced(e.g.,more ruminating than walking)resulting in u...Classification of sheep behaviour from a sequence of tri-axial accelerometer data has the potential to enhance sheep management.Sheep behaviour is inherently imbalanced(e.g.,more ruminating than walking)resulting in underperforming classification for the minority activities which hold importance.Existing works have not addressed class imbalance and use traditional machine learning techniques,e.g.,Random Forest(RF).We investigated Deep Learning(DL)models,namely,Long Short Term Memory(LSTM)and Bidirectional LSTM(BLSTM),appropriate for sequential data,from imbalanced data.Two data sets were collected in normal grazing conditions using jaw-mounted and earmounted sensors.Novel to this study,alongside typical single classes,e.g.,walking,depending on the behaviours,data samples were labelled with compound classes,e.g.,walking_-grazing.The number of steps a sheep performed in the observed 10 s time window was also recorded and incorporated in the models.We designed several multi-class classification studies with imbalance being addressed using synthetic data.DL models achieved superior performance to traditional ML models,especially with augmented data(e.g.,4-Class+Steps:LSTM 88.0%,RF 82.5%).DL methods showed superior generalisability on unseen sheep(i.e.,F1-score:BLSTM 0.84,LSTM 0.83,RF 0.65).LSTM,BLSTM and RF achieved sub-millisecond average inference time,making them suitable for real-time applications.The results demonstrate the effectiveness of DL models for sheep behaviour classification in grazing conditions.The results also demonstrate the DL techniques can generalise across different sheep.The study presents a strong foundation of the development of such models for real-time animal monitoring.展开更多
基金The authors would like to acknowledge Alexander Clancy and Jaime Manning for their assistance in establishing this trial and Dougal Pottie for his assistance in the field.Financial support for this trial was provided by the Australian Wool Education Trust grant to Alexander Clancy and the Coolringdon Research Trust,which provides a scholarship to Danica Parnell.
文摘Background:Grasslands are the primary source of feed for grazing livestock,and as such,knowledge on how to best manage livestock and grasslands,through the use of spatiotemporal modelling,will assist in the long-term management of a valuable ecosystem resource.Methods:This study was conducted over 14 months between March and April 2017 in Orange,NSW,Australia.The study evaluated sheep behaviour in relation to the presence of pasture species,environment and paddock structures,using random forest modelling,to predict sheep location under continuous high(HSR,13 DSE ha−1)and low(LSR,7DSE ha−1)stocking rates.Results:In the LSR,significant drivers included water,shade and fence lines(p<0.01).In the HSR,only fence lines and available biomass were found to be significant(p<0.01).The presence of green legumes in both stocking rates often increased residency by sheep.Animals spent more time together in the LSR,suggesting that social behaviour played a larger role than pasture quantity and quality in driving grazing behaviours.Conclusions:Understanding how pasture type can influence grazing behaviours and also how animal behaviour affects pasture performance and utilisation is important in developing long-term sustainable management strategies on a paddock scale.
文摘Classification of sheep behaviour from a sequence of tri-axial accelerometer data has the potential to enhance sheep management.Sheep behaviour is inherently imbalanced(e.g.,more ruminating than walking)resulting in underperforming classification for the minority activities which hold importance.Existing works have not addressed class imbalance and use traditional machine learning techniques,e.g.,Random Forest(RF).We investigated Deep Learning(DL)models,namely,Long Short Term Memory(LSTM)and Bidirectional LSTM(BLSTM),appropriate for sequential data,from imbalanced data.Two data sets were collected in normal grazing conditions using jaw-mounted and earmounted sensors.Novel to this study,alongside typical single classes,e.g.,walking,depending on the behaviours,data samples were labelled with compound classes,e.g.,walking_-grazing.The number of steps a sheep performed in the observed 10 s time window was also recorded and incorporated in the models.We designed several multi-class classification studies with imbalance being addressed using synthetic data.DL models achieved superior performance to traditional ML models,especially with augmented data(e.g.,4-Class+Steps:LSTM 88.0%,RF 82.5%).DL methods showed superior generalisability on unseen sheep(i.e.,F1-score:BLSTM 0.84,LSTM 0.83,RF 0.65).LSTM,BLSTM and RF achieved sub-millisecond average inference time,making them suitable for real-time applications.The results demonstrate the effectiveness of DL models for sheep behaviour classification in grazing conditions.The results also demonstrate the DL techniques can generalise across different sheep.The study presents a strong foundation of the development of such models for real-time animal monitoring.
