Using automated image analysis in pig behavioural research: Assessment of the influence of enrichment substrate provision on lying behaviour

Visual monitoring of pig behaviours over long periods is very time consuming and has possibility for observer bias. Automated image processing techniques now give the potential to carry out behavioural research in a more effective way. To illustrate this, an image processing technique was applied to identify whether any changes in pig lying behaviour which might be detrimental to welfare resulted from an enrichment provision treatment. The lying patterns of pigs in 6 enriched pens were compared with those of 6 control pens, which had only a suspended enrichment toy, to determine whether daily provision of a rooting material (maize silage) onto a solid plate in the lying area of a fully slatted pen resulted in changed lying time and location. Pigs were monitored by top view CCTV cameras and animals were extracted from their background using image processing algorithms. An ellipse fitting technique was applied to localize each pig and the centre of each fitted ellipse was used in x-y coordinates to find the lying positions after use of an algorithm to remove images in motion preceding the scan. Each pen was virtually subdivided into four zones and the position of each lying pig obtained at 10 min intervals over a series of 24 h periods. Results of a validation study showed that the image processing technique had an accuracy of 93-95% when compared to visual scoring. Results from image processing indicated that once daily provision of rooting material significantly changed the diurnal activity pattern (p < 0.001) and resulted in a modified diurnal pattern of resting location. The study demonstrates that machine vision can be used as a precise and rapid method for quantifying pig lying behaviour for research or practical applications.