Recent research advances, challenges and trends in non-invasive infrared thermography for the detection of mastitis in dairy cows

Mastitis is a highly prevalent and dangerous disease that not only has a serious impact on milk production and quality of cows, but also poses a threat to human health. Non-invasive detection of udder health status is a key tool to safeguard animal welfare and improve the efficiency of farm operations. However, the lack of standardised standard-operating procedures (SOPs) and inter-laboratory calibration protocols still limits the large-scale deployment of infrared thermography (IRT)-based mastitis detection. This paper reviews the results of the past twenty-five years of research on non-invasive dairy cow mastitis detection using infrared thermography technology. First, the key factors affecting the accuracy of temperature measurement are systematically clarified, including device parameters, environmental conditions, milking methods, and data acquisition procedures. Second, the mechanisms by which region of Interest labelling strategies, the robustness of intelligent recognition algorithms, and the choice of temperature statistics influence detection performance are summarised. Third, it is discussed that cut-off values-based and threshold-based classification methods are vulnerable to individual heterogeneity and scenario shift, thereby weakening the capability and stability of mastitis identification. Finally, this review proposes a multimodal fusion framework based on augmented reality (AR) glasses and their supporting system for future multi-scenario detection, integrating multisource data acquisition, on-device inference, quality self-checks, external validation, drift monitoring, model updates, and graded intervention feedback, thereby supporting more standardised and scalable contactless early warning in smart barns.