| Abstract |
Subclinical and clinical disease is the biggest factor responsible for pig system inefficiency and reductions in productivity and welfare. Currently disease or vice detection is done either via human observation or using diagnostic surveillance, both of which have limitations w.r.t. cost and effort required for continuous, high-frequency monitoring of large numbers of animals. This project aims to develop and validate innovative technology to automatically monitor the performance and behaviour in grower and finisher pigs systems, with the objective of automatically detecting the consequences of health and welfare challenges on farm. Through automation we will enable continuous and objective analysis of animal well being. Combined with innovative analysis methods this serves as the basis for early detection of the consequences of health and welfare challenges, and thus for rapid intervention that will lead to increased efficiency on farms. Our technical approach will comprise computer vision and pattern recognition methods for monitoring groups of pigs in indoor pens. We will refine an existing visual analysis system that estimates liveweights at feeders, towards continuous, and location independent analysis. Furthermore, our visual monitoring system will track locations and dynamics of pig movements. Based on these continuous visual observations we will develop methods for modelling normal, i.e., expected performance development and behaviour, and for detecting deviations from this 'normality'. These analysis techniques will be calibrated such that they abstract from environmental factors such as temperature and humidity. An automated early warning system will lead to: i) earlier detection of health & welfare issues enabling effective intervention, which is grounded in the academic partner's work demonstrating that behaviour changes manifest long before clinical disease signs; ii) increased efficiency and reduce costs, especially for large-scale operations. The project will thus contribute towards sustainability and competitiveness of the UK pig Industry. The technical approach of the proposed project will: i) Develop a robust, camera-based monitoring system for the analysis of both behaviour and performance development in pigs that is suitable for continuous monitoring of groups of animals; ii) Develop algorithms that model normal behaviour of individuals and groups of animals, and allow for quantitative measurements of relevant development criteria and develop automatic assessment methods that detect deviations from normal, i.e., expected development and behaviour, during controlled and spontaneous health and welfare problems; iii) Implement the proposed system within a cloud- and mobile computing infrastructure for wide accessibility and near real-time feedback to farm personnel; iv) Validate the framework in realistic deployments as a means of detecting the onset of health/welfare problems on pig farms; and (v) Ensure effective KT to the relevant stakeholders. . The project brings together the UK leading designer of innovative software solutions for the agricultural sector (Innovent), the world's leading animal health company (Zoetis) and two of the UK's leading companies for pig health and management (Raft and Harbro), with a UK University that is at the forefront of research in computer vision, pattern recognition techniques and pig management and health (Newcastle University). Additional funding from the British Pig Executive (BPEX) will ensure that the outcomes of the project will be relevant and disseminated to the wider UK pig industry. |
