Pig Welfare Through Behavior Learning From Camera Recordings

Animal Science students along with the farm staff have to monitor the behavior of pigs in order to assure their welfare. The video systems are used by our educational software and new methods of pig observation, evaluation and treatment are applied much faster and more efficient compared to the classical intervention. Each recording is stored as a media file and each frame taken at 0.1 seconds is stored as a Bitmap image. The Bitmap images are processed in parallel using the MapReduce programming model from Apache Hadoop. The contour of the image is automatically analyzed and based on it the presence of pigs is detected, as well as their location can be determined. The location is important because it can be denoting that the pig eats or that it stays aside. Pig limp was also detected. It was observed based on the recordings that 83% of the time the pigs spend it lying down, 7% is spent eating and 10% of the time they walk and sit. Video monitoring and automatic interpretation facilitates the learning of new intervention approaches and boosts the responsiveness among the students. The students can learn from the critical situations and benefit from these cases while learning.