Please use this identifier to cite or link to this item: https://hdl.handle.net/1959.11/26563
Title: Comparison of grazing behaviour of sheep on pasture with different sward surface heights using an inertial measurement unit sensor
Contributor(s): Guo, Leifeng (author); Welch, Mitchell  (author)orcid ; Dobos, Robin  (author)orcid ; Kwan, Paul  (author); Wang, Wensheng (author)
Publication Date: 2018-07
Early Online Version: 2018-05-22
DOI: 10.1016/j.compag.2018.05.004
Handle Link: https://hdl.handle.net/1959.11/26563
Abstract: Grazing is the most important activity that ruminant livestock undertake daily. A number of studies have used motion sensors to study the grazing behaviour of ruminant livestock. However, few have attempted to validate their approaches against various sward surface heights (SSH). The objectives of our study were to: (1) identify and compare the effects of different SSH on the grazing behaviour of sheep by analyzing data collected by a collar mounted Inertial Measurement Unit (IMU) sensor; (2) calculate the relative importance of the extracted features on grazing identification and compare the consistency of the selected features across various SSH; (3) validate the robustness by using classifiers trained from the dataset with specific SSH to distinguish the grazing activity on the datasets from different SSH; and (4) visualize the classification results of grazing versus non-grazing activities on various SSH. Linear Discriminant Analysis (LDA) was chosen as the classification method, while Probabilistic Principal Component Analysis (PPCA) was used to reduce dimensionality of the feature space for visualization of the results. Experimental results revealed that (1) our approach achieved high classification accuracy of grazing behaviour (over 95%) on all the epochs regardless of SSH; (2) Mean of accelerometer Z-axis, Entropy of accelerometer Y-axis, Entropy of accelerometer Z-axis, Mean of gyroscope X-axis and Mean of gyroscope Y-axis were the top 5 features that contributed most in classifying the grazing versus non-grazing activities and there were consistent trends in features across the three SSH; (3) there was enough robustness when the trained LDA classifier on a specific SSH was used to classify behaviour on different SSH; and (4) there existed a clear linear boundary between the data points representing grazing and those of non-grazing behaviour. Overall, our research confirmed that IMU sensors can be a very effective tool for identifying the grazing behaviour of sheep and there is enough robustness to use a trained LDA classifier on a specific pasture SSH to classify grazing behaviour at different SSH pastures.
Publication Type: Journal Article
Source of Publication: Computers and Electronics in Agriculture, v.150, p. 394-401
Publisher: Elsevier BV
Place of Publication: The Netherlands
ISSN: 0168-1699
1872-7107
Field of Research (FOR): 080109 Pattern Recognition and Data Mining
070203 Animal Management
Socio-Economic Outcome Codes: 830310 Sheep - Meat
830311 Sheep - Wool
970107 Expanding Knowledge in the Agricultural and Veterinary Sciences
Peer Reviewed: Yes
HERDC Category Description: C1 Refereed Article in a Scholarly Journal
Appears in Collections:Journal Article
School of Science and Technology

Files in This Item:
1 files
File SizeFormat 
Show full item record
Google Media

Google ScholarTM

Check

Altmetric


Items in Research UNE are protected by copyright, with all rights reserved, unless otherwise indicated.