Change in cows behaviours is one of the indicators useful to help identifying when animals become ill. The need to analyse a large number of animals at a time due to the increase in the herd dimension in intensive farming has led to the use of automated systems. Among automated systems, inertial sensor-based systems have been utilised to distinguish behavioural patterns in livestock animals. In this field, the overall aim of this thesis work, which was inherent to the field of the Precision Livestock Farming, was to contribute to the improvement of the systems based on wearable sensors that are able to recognise the main behavioural activities (i.e., lying, standing, feeding, and walking) of dairy cows housed in a free-stall barn. This objective was achieved through different steps aimed at producing an advance in the state of the art. A novel algorithm, characterised by a linear computational time, was implemented with the aim to improve real-time monitoring and analysis of walking behaviour of dairy cows. The algorithm computed the number of steps of each cow from accelerometer data by making use of statistically defined thresholds. Algorithm accuracy was carried out by computing total error (E equal to 9.5 %) and Relative Measurement Error (RME between 2.4% and 4.8%). A new classifier was assessed to recognise the cow feeding and standing behavioural activities by using statistically defined thresholds computed from accelerometer data. The accuracy of the classification was assessed by computing of the Misclassification Rate (MR equal to 5.56%). A new data acquisition system assessed in a free-stall barn allowed the acquisition of data from different sensor devices, with a sampling frequency of 4 Hz, during the animals daily routine. It required a simple installation into the building and it did not need any preliminary calibration. The performance of this system was assessed by computing a Stored Data Index (DSI) that resulted equal to 83%. Finally, the overall design of an automated monitoring system based on wearable sensors was proposed.
L' alterazione del comportamento degli animali è uno degli indicatori utili per identificare l insorgenza di malattie. La necessità di controllare un numero sempre maggiore di capi negli allevamenti intensivi ha portato all utilizzo di sistemi automatizzati per il loro monitoraggio. Tra questi, i sistemi basati su sensori inerziali sono stati recentemente proposti per classificare i pattern comportamentali degli animali negli allevamenti. In questo ambito, che è inerente al campo della Precision Livestock Farming, il lavoro svolto durante il Dottorato di ricerca e descritto nella presente tesi si propone di contribuire al miglioramento di tali sistemi per il riconoscimento delle attività di lying, standing, feeding e walking delle bovine da latte allevate in una stalla a stabulazione libera. Sulla base di un ampia analisi dello stato dell arte, tale obiettivo è stato conseguito tramite la definizione di nuovi approcci di applicazione della ICT (Information and Communications Technology) alla zootecnia intensiva. In particolare, è stato realizzato un nuovo algoritmo, caratterizzato da un complessità computazionale lineare, che effettua il calcolo del numero di passi di ogni bovina dai dati di accelerazione, facendo uso di soglie definite statisticamente. L accuratezza dell algoritmo è stata valutata sulla base dell errore totale, pari al 9.5%, e del Relative Measurement Error, compreso tra il 2.4% e il 4.8%. Inoltre, è stato definito un nuovo classificatore per distinguere l attività del feeding dallo standing, utilizzando soglie calcolate statisticamente dai dati accelerometrici. L accuratezza della classificazione è stata valutata sulla base del Misclassification Rate, pari al 5.56%. L applicazione in stalla di un nuovo sistema di acquisizione dei dati ha permesso di migliorare la raccolta dei dati da differenti sensori durante la routine giornaliera degli animali. Il sistema proposto richiede un installazione facilitata e non necessita di calibrazioni preliminari. La sua prestazione è stata valutata mediante lo Stored Data Index che è risultato pari all 83%. Infine, viene proposto il progetto di un sistema complessivo per il monitoraggio in automatico dei comportamenti delle bovine basato su sensori indossabili.
Design of an automated system for continuous monitoring of dairy cow behaviour in free-stall barns / Mancino, Massimo. - (2017 Jan 30).
Design of an automated system for continuous monitoring of dairy cow behaviour in free-stall barns
MANCINO, MASSIMO
2017-01-30
Abstract
Change in cows behaviours is one of the indicators useful to help identifying when animals become ill. The need to analyse a large number of animals at a time due to the increase in the herd dimension in intensive farming has led to the use of automated systems. Among automated systems, inertial sensor-based systems have been utilised to distinguish behavioural patterns in livestock animals. In this field, the overall aim of this thesis work, which was inherent to the field of the Precision Livestock Farming, was to contribute to the improvement of the systems based on wearable sensors that are able to recognise the main behavioural activities (i.e., lying, standing, feeding, and walking) of dairy cows housed in a free-stall barn. This objective was achieved through different steps aimed at producing an advance in the state of the art. A novel algorithm, characterised by a linear computational time, was implemented with the aim to improve real-time monitoring and analysis of walking behaviour of dairy cows. The algorithm computed the number of steps of each cow from accelerometer data by making use of statistically defined thresholds. Algorithm accuracy was carried out by computing total error (E equal to 9.5 %) and Relative Measurement Error (RME between 2.4% and 4.8%). A new classifier was assessed to recognise the cow feeding and standing behavioural activities by using statistically defined thresholds computed from accelerometer data. The accuracy of the classification was assessed by computing of the Misclassification Rate (MR equal to 5.56%). A new data acquisition system assessed in a free-stall barn allowed the acquisition of data from different sensor devices, with a sampling frequency of 4 Hz, during the animals daily routine. It required a simple installation into the building and it did not need any preliminary calibration. The performance of this system was assessed by computing a Stored Data Index (DSI) that resulted equal to 83%. Finally, the overall design of an automated monitoring system based on wearable sensors was proposed.File | Dimensione | Formato | |
---|---|---|---|
PhD Thesis MMancino.pdf
accesso aperto
Tipologia:
Tesi di dottorato
Licenza:
PUBBLICO - Pubblico con Copyright
Dimensione
1.48 MB
Formato
Adobe PDF
|
1.48 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.