Automated recognition of postures for the detection of compromised health pig

Alameer, A ORCID: https://orcid.org/0000-0002-7969-3609 'Automated recognition of postures for the detection of compromised health pig' , in: 71st Annual Meeting of European Federation of Animal Science , http://www.eaap.org/EAAP2020_Book_of_Abstracts.pdf.

[img] PDF - Published Version
Restricted to Repository staff only

Download (20MB)

Abstract

Changes in pig behaviours may be used to detect early signs of problems, such as in animal health. Automated vision-based early warning systems have been developed to detect behavioural changes in groups of pigs to monitor their health and welfare status. In commercial settings, automatic detection of pig postures and drinking behaviour remains a challenge. Here, we developed a system that automatically identifies pig postures (standing, sitting, lying lateral and lying sternal) and drinking behaviour. Pigs were monitored by top view RGB cameras that covered a large area of the pen including the drinking area, and animal behaviours were detected using deep learning-based methods. Our first objective was to demonstrate the ability of this automated method to identify behaviours of individual animals with high precision. We then tested the system ability to detect changes in group-level behaviours due to a food restriction protocol. Two deep learning-based detector methods, including faster regions with convolutional neural network features (Faster R-CNN) and you only look once (YOLO) combined with Residual Network (ResNet-50), were developed to precisely identify pig postures and drinking behaviours of group-housed pigs. We evaluated our method using routine data recorded at a commercial pig farming environment. Our experiments show that our system could recognise the postures and drinking behaviour of individual pigs with a mean average precision (mAP) of 0.9888 ± 0.0094. When the pig feeding regime was disrupted, we observed significant deviations from the daily ad-libitum routine in the standing, lateral lying and drinking behavoiurs. These experiments demonstrate this method is capable of robustly and accurately monitoring pig behavoiurs under commercial conditions without the need for additional sensors or individual markings.

Item Type: Book Section
Schools: Schools > School of Computing, Science and Engineering
Journal or Publication Title: 71st Annual Meeting of European Federation of Animal Science
Publisher: http://www.eaap.org/EAAP2020_Book_of_Abstracts.pdf
Depositing User: A Alameer
Date Deposited: 27 May 2022 11:56
Last Modified: 17 Aug 2022 09:33
URI: https://usir.salford.ac.uk/id/eprint/63694

Actions (login required)

Edit record (repository staff only) Edit record (repository staff only)