Teat detection mechanism using machine learning based vision for smart Automatic Milking Systems

This paper discusses the work in progress of Automatic Milking system with focus on development of a smart vision system for the manipulator of the milking robot. The vision system being developed aims at giving the manipulator the capability for faster and accurate teat detection. In our laboratory environment, we are developing a smart vision system which is trained by machine learning and can detect the cow's udder and the teats for all the pre-milking, milking and post-milking processes with least to none time being spent on cow herd training for the system in the barn environment. This teat detection mechanism aims to reduce the time spent on training of the system in the first few initial milkings and reduce operator responsibilities, to eventually produce a more of plug and play product.