Browse by author
Lookup NU author(s): Dr Ali Alameer,
Dr Hillary Dalton,
Professor Ilias Kyriazakis
Full text for this publication is not currently held within this repository. Alternative links are provided below where available.
Automated vision-based early warning systems have been developed to detect behavioural changes in groups of pigs to monitor their health and welfare status. However, automatic feed detection remains a problem in precision pig farming due to problems of light alteration, occlusions and the similar appearances of pigs. Additionally, these systems often overestimate the actual time spent feeding due to the inability to identify and exclude non-nutritive visits (NNV) to the feeding area. To tackle these problems, we developed a robust feed-detection method that is capable of distinguishing between feeding and NNV to the feeding area for group-housed pigs. Our first objective was to demonstrate the ability of this automated method to identify feeding and NNV behaviour with high accuracy. We then tested the system’s ability to detect a disturbance in group-level feeding and NNV behaviour due to feed restriction. A GoogLeNet deep learning model was utilised as a base network to accurately identify the feeding (i.e., pig has head inside the feeding trough) and NNV (i.e., pig enters the black mat area with two or more feet with one being a front foot) behaviour of group-housed pigs. The method was designed to monitor a predefined pen area covering two feed troughs and a black mat covering the area in front of feeders using a video surveillance camera. The experimental tests showed that our system could recognise the feeding and NNV behaviour of pigs with an accuracy of 99.4%. Following this validation, we tested the method’s ability to detect group level feeding and NNV changes from a normal ad libitum feeding date and a date of quantitive feed restriction. These experiments demonstrate this method is capable of robustly and accurately monitoring the feeding behaviour of groups of pigs under commercial conditions without the need for additional sensors or individual marking.
Author(s): Alameer A, Dalton H, Barcardit J, Kyriazakis I
Publication type: Conference Proceedings (inc. Abstract)
Publication status: Published
Conference Name: 70th Annual Meeting of the European Federation of Animal Science
Year of Conference: 2019
Print publication date: 30/08/2019
Acceptance date: 15/04/2019
Publisher: Wageningen Academic Publishers
Library holdings: Search Newcastle University Library for this item