Pig video tracking methods: a review and new trends

Gang Jun Tu & Erik Jørgensen

Poster

65th Annual Meeting of the European Federation of Animal Science

Abstract

Light changes (e.g. switch on/off in the pig house), dynamic background (e.g. straw in the pig pen) and motionless foreground (e.g. pigs don’t move) are three main factors that are always encountered in complex pig pens. These factors seriously affect the process of pig video tracking. The objective of this paper is to review existing pig video tracking methods with the aim of revealing the strengths and weaknesses of the methodologies based on the above three problems, and introduce the new trends of methodological development. In order to develop effective solutions to these important issues in pig video tracking, we
give an overview of the processing framework of pig tracking in general: image preprocessing, object segmentation or object modelling, object tracking. Then, we categorize the existing methods into two groups (i.e. based on object segmentation or object modelling), provide detailed descriptions of representative methods in each group. The potential in resolving the three major obstacles above discussed has been examined for each method, and this analysis suggests that methods based on wavelet transform exhibit a good potential to overcome the light changes, dynamic background and motionless foreground in the complex pen, since local texture difference measures between a current image and a reference image can be calculated by using the wavelet gradient that are less sensitive to illumination changes, and the dynamic background can be regularized to a relatively static background by using wavelet decomposition. Our research demonstrates that wavelet transform has great potential in pig tracking as it provides a rich representation of a pig. In conclusion, the comparative analyses of existing video tracking methods presented in this review will help the researchers to fulfil their monitoring pig behaviour in complex pig pens and help to further optimize wavelet transform-based algorithms for their application in video tracking of animal behaviour.

[Poster]