Designing AI-Based Non-invasive Method for Automatic Detection of Bovine Mastitis

Mastitis is a major disease in dairy animals as a consequence of udder inflammation. The occurrence of disease in dairy animals impacts farmers economic status and hinders dairy industry growth. Many conventional and rapid methods are used by the dairy industry to detect mastitis by considering the moderate to severe symptoms of infected animals and their milk. The limitations of such methods are that they are expensive, arduous, and require samples. In the precision dairying era, rapid, low-cost, automated alternative techniques for early prediction of diseases in animals are in demand by stakeholders, especially farmers. Infrared Thermography (IRT) is an emerging, non-invasive tool to predict diseases in humans and animals. IRT, coupled with machine learning algorithms, has the potential to detect bovine mastitis. Researchers explored algorithms such as K-nearest neighbourhood (KNN), Support vector machine (SVM), Random Forest (RF), and Convolutional Neural Network (CNN) for automatic, real-time detection of mastitis using thermographic images. This paper discusses the prevalence of mastitis, related works using IRT with machine learning models, designing a model by embedding domain heuristics to fine-tune the decision, and details about the experiment carried out by employing KNN and SVM for thermal and demographic data. Both SVM and KNN classifiers classified the disease with an accuracy of 60%; this low accuracy is due to limited data. However, the results provide new insights to develop a non-invasive method for the detection of mastitis. The future work envisaged by using large field data for the detection of mastitis more accurately.