Robotic application of milking cups to the udder of a cow in a rotary high capacity group milking system is a major challenge in automated milking. Application time and reliability are the main constraints. Manual application by an operator of a rotary carousel is of the order of 10 seconds and 100% reliable. In existing non-rotary milking machines, the cups are applied to each teat individually and the process can take up to two minutes. In order to achieve a more rapid simultaneous application of the four cups, the three dimensional locations of the four teats must be known in real time. In this thesis, a multimodal vision system combining optical stereovision and thermal imaging is developed. The overall system is evaluated from the point of view of accuracy and robustness. Laboratory tests have shown that stereovision can rapidly locate teat three dimensional position coordinates, however robust identification of the teats is required. It is shown that this may be achieved using thermal imaging to isolate teats from background objects due to their elevated temperature profile. Further development is necessary to overcome specific situations such as overlapping teats.