Computers and electronics in agriculture vol:20 issue:1 pages:21-29
Oestrus detection of sows is an important activity within pig husbandry. It takes about 30% of the overall labour input and financial losses related to non-productive days may vary considerably. Therefore, the principle of using an infra-red sensor has been evaluated to detect automatically oestrus of sows. A total of 58 multiparous individually housed sows were monitored from the day after the piglets were weaned. The infra-red sensor was mounted 50 cm above the front part of the sow's body. Four parameters of body movement as quantified by the sensor's output voltage were investigated: mean daily activity; standard deviation of mean daily activity; minimal; and peak value. The reference method was the standing behaviour before the boar and the inseminator. Four classes of behaviour were defined: (1) standing + increase of mean daily activity; (2) standing + no increase; (3) no standing + no increase; and (4) no standing + increase. Up to 80% of the sows could be classified correctly when using mean daily activity as the selection parameter. Up to 86% were classified correctly when daily peak activity was also included. Positive and false positive sows could be distinguished at the 95% level by using a canonical discriminant analysis. Increase of mean daily activity, increase of standard deviation and increase of daily peak activity were statistically significant explanatory variables of a legit function predicting the four classes of oestrous behaviour. (C) 1998 Elsevier Science B.V. All rights reserved.