Combination of image and sound analysis for behaviour monitoring in pigs

Welfare assessment of livestock takes a lot of time nowadays. Automatic, continuous monitoring of welfare measures would make a substantial contribution to the implementation of welfare assessment on the farm. The purpose of this work was to investigate a combination of sound and image analysis techniques for monitoring livestock behaviour which was relevant to the assessment of animal welfare. The non-invasive nature of vision and sound technology makes it very attractive for many monitoring applications in livestock husbandry. This study comprised measurements on four groups of piglets, with six piglets per group in a pen. Image and sound recordings were taken on 13 days with 12 hours of recording from 7h to 19h. Each pen was monitored by a top-view CCD camera and 3 microphones installed around the pen. Image processing techniques were used to calculate the difference in pixel intensity between frames in order to measure image activity. The sound analysis involved measuring sound energy together with noise reduction in order to measure the energy of vocalisations. Results for activity and vocalisation energy were then compared. The data demonstrated that there is an 80% correlation between activity and vocalisation in pigs.