Image analysis to predict the moment of giving birth of pregnant ewes

A. Chedad, J.M. Aerts, A. Delahaye and D. Berckmans

Laboratory for Agricultural Buildings Research, University of Leuven, Heverlee, Belgium

Due to the evolution of hardware and software, computers become more and more reliable, cheap and powerful. As a consequence digital image analysis techniques are used more and more integrated in bio-processes [2]. Such techniques are accurate, no contact has to be made and the continuous presence of man is not necessary. The objective of this research was to investigate if image analysis techniques can be used to quantify the behaviour of pregnant ewes in field conditions.

Figure 1. Test installation.

An algorithm for image analysis was developed which allows making a distinction between lying and standing behaviour of ewes and this in situations of low contrast between animal and background. Using a monochrome CCD camera (speed of 25 images/second) and video recorder, the behaviour of four ewes (3 pregnant ewes and 1 non-pregnant ewe) was recorded starting 2 to 6 days before having a litter, depending on the ewe.

Figure 2. Cumulative traveled distance for ewe 2.

For a total of 34727 digitised images it was possible for 95 % of the images to make a correct distinction between lying and standing and this in lighting conditions of only 50 lux.

Figure 3. Cumulative lying Behaviour for ewe 2.

The following variables could be quantified: travelled distance (m), lying time (seconds) , standing time (seconds), activity (defined by [1]) and rotation index (number of rotations) of an animal. For 2 of the 3 pregnant ewes a change in behaviour could be quantified 6 to 7 hours before birth. In Figures 2 and 3 the traveled distance and the cumulative lying behaviour is shown for ewe 2.

Using such image analysis techniques it might be possible to monitor on-line the behaviour of pregnant animals in order to detect the beginning of birth.


  1. Bloemen, H.; Aerts, J.M.; Berckmans, D.; Goedseels, V. (1997). Image analysis to measure activity of animals. Equine Veterinary Journal, Suppl. 23, 16-19.
  2. Deschazer, J.A.; Moran, P.; Onyyango, C.M.; Randall, J.M.; Schofield, C.P. (1988). Image Systems to Improve Stockmanship in Pig Production. Divisional Note 1459 DN, AFRC Inst. Engineering Research, Silsoe.

Poster presented at Measuring Behavior 2000, 3rd International Conference on Methods and Techniques in Behavioral Research, 15-18 August 2000, Nijmegen, The Netherlands

© 2000 Noldus Information Technology b.v.