Automated video analysis of pig activity at pen level highly correlates to human observations of behavioural activities

Automated collection of continuous activity data of pigs can be performed easily using video analysis. In welfare and health research, this technique can be economically advantageous over manual observations. However, the relationship between activity measures by automated video analysis and manually scored behavioural activity has never been established. We correlated automated activity measures through video analysis to ethological scores of pig activity, using off-line video recordings of four pens with grower pigs. Human observations (HO) of different behavioural activities were carried out by 2-min scan sampling during four 30-min sessions on 6 observation days. HO of pig activity was expressed as a mean proportion per session. Automated observations (AO) of pig activity were calculated by the relative number of moving pixels between two consecutive image frames (1 frame/second) and expressed as a mean image activity index per session. The overall correlation between pig activity data from AO and HO was strong and positive (Rs=0.92, P<0.0001). When comparing AO and HO data at session level, the correlation coefficients for the two afternoon sessions were lower. Both static activities and activities involving locomotion had a significant effect on the activity index of AO (P