Automated equine behaviour monitoring

点击翻译按钮获取中文摘要

The astute caretaker and the veterinarian heavily rely on the assessing unusual behaviour or a change in behaviour as the first sign of physical pain or disease. Stereotypies are repetitive, highly stylized, seemingly functionless motor responses and sequences that horses can develop and can signifiy abnormal behaviour, misbehaviour or coping mechanism. The experts (such as the caretaker or vet) will identify stereotypies as possible disease symptoms after management conditions (i.e., housing, social, exercise, nutrition) and genetic predisposition are already considered. Sport horses are prone to several potentially serious ailments from leg injuries to colic that can lead to sudden death if not treated shortly after symptoms occur. Currently, watching sport horses in person, or by web cam, are the only behaviour monitoring solutions available. This is effective when constant monitoring is complied to but this is not sustainable and unusual behaviours many times go undetected. An automatic monitoring technique would help in overcoming these limitations. This project will investigate using computer vision techniques to automate the process of flagging unusual behaviour in the horse and sending the video sequence to the horse owner, vet or caretaker. For example, colic in horses is associated with abdominal pain and abdominal pain is exhibited when the horse bites at his sides, lays down for longer than normal, continually shifts his weight on the hind limbs, paws or has a described appetite or does not eat at all. Severe abdominal pain includes signs of profuse sweating, continuous rolling, persistent movement and getting up and down violently. Computer vision techniques that will be explored include feature tracking, optical flow as well as skeletal action recognition techniques. The anticipated result of this project is an assessment of what behaviours can be detected with what techniques and at what accuracy.