Usage of computer vision analysis for automatic detection of activity changes in sows during final gestation

摘要

Surveillance of farm animals and automatic detection of deviant behaviors has lately become more important in livestock science and farming. With increasing farm size and diversity in tasks, farmers can benefit from automatic animal behavioral surveillance. This application note describes the usage of two straightforward computer vision algorithms with the purpose to analyze and record the movement activity of single-housed sows. The entire process from data acquisition over pre- and post-processing is provided. Additionally, the obtained data is used for automatic prediction of imminent farrowing in sows with sow individual CUSUM-Charts.