Video tracking of dairy cows for assessing mobility scores

Abstract

Lameness afflicts a large proportion of dairy herds, but could be considerably reduced by automated monitoring by CCTV. Key to this is reliable, robust detection and tracking of individual cows in crowded video sequences. We introduce a novel detection and tracking method, based on the Viola-Jones detector. We show that animals can be tracked and their overall gait patterns and speed automatically extracted from video sequences. Preliminary work on identification of individual animals through principal component analysis and SIFT feature matching is also described.