Objective Tracking of Gait Patterns in Trotters

This thesis proposes an objective foundation for analysing the gait patterns oftrotting racehorses using computer vision techniques, with a particular focus onidentifying gait asymmetries. The use of a Minimum Output Sum of SquaredError (MOSSE) tracker and a stereo camera system enhances the accuracy androbustness of tracking in dynamic environments. The dataset, provided by theSwedish University of Agricultural Sciences (SLU), includes videos captured un-der varying conditions using both monocular and stereo cameras. Key investi-gations address the role of image features in improving tracking effectiveness,the advantages of stereo vision over monocular setups, and the influence of fea-ture selection, video stabilisation and frame rate on tracking performance. Find-ings indicate that integrating stereo camera data and advanced image featuressignificantly improves tracking robustness, offering a reliable and objective pathforward for detecting lameness in trotting racehorses. The tested method hasthe potential to enhance equine welfare through early diagnosis and interventionwhile advancing veterinary and computer vision applications.