奶牛乳房炎自动检测技术研究进展

奶牛乳房炎是影响奶牛健康的主要疾病之一，发病率高、发病范围广、经济损失严重。目前奶牛乳房炎的检测大多是采集奶牛乳汁进行理化性质检测，该方法对检测环境有着较高要求，且检测周期长。随着信息技术的迅速发展，奶牛乳房炎的自动检测技术取得了较好的研究成果。该研究根据数据的传感器类型，从视觉传感器、自动挤奶系统与其他传感器3个方面阐述了奶牛乳房炎自动检测的研究进展。基于视觉传感器的奶牛乳房炎自动检测方法包括基于乳房热红外图像和基于乳汁图像的检测方法，该方法较大程度上保障了动物福利，但检测精度有待提升；基于自动挤奶系统（automatic milking systems, AMS）的奶牛乳房炎自动检测方法利用AMS获取乳汁信息，然后构建乳房炎检测模型，该方法检测误差较小，但成本较高；基于其他传感器的奶牛乳房炎检测方法采用独立研发的传感器获取乳汁数据，预测乳房炎发病情况，该方法操作简便，但使用不同传感器构建的检测模型精度差异较大。该文还探讨了目前奶牛乳房炎自动检测领域存在的精度低、实时性差、自动化不足等问题，并展望了该领域未来的发展趋势，以期为开展奶牛乳房炎自动检测技术与方法研究提供参考。

Research advances in the automatic detection technology for mastitis of dairy cows

Cow mastitis is one of the most serious diseases in the healthy development of dairy farming, due to the high incidence and wide range of characteristics. Mastitis in dairy cows can also reduce milk production and quality, leading to human health and herd turnover costs. Dairy cows with clinical mastitis vary greatly in the abnormal changes in their udders and milk. In the case of subclinical mastitis, there is no outstanding appearance in the udder and milk, where the economic loss is more severe than before. Physical and chemical property testing has been widely used to collect the milk for the diagnosis of mastitis in dairy cows at present. However, the testing environment and the long testing period cannot fully meet the demand for the rapid and real-time diagnosis of mastitis in dairy cows. Fortunately, advanced electronic information technology and equipment have been widely used in the field of agriculture in recent years. The automatic detection of dairy cow mastitis has achieved better research for the high requirements. In this review, the current research progress was introduced into the automatic detection of dairy cow mastitis, according to the different types of sensors in the data acquisition. Three aspects were also included: the visual sensor, the automatic milking system (AMS), and the rest sensors. Firstly, the automatic detection of dairy cow mastitis with the vision sensor was divided into the detection with the udder thermal infrared and the milk images. Automatic detection was achieved in mastitis in dairy cows without damage and stress. Especially, the mastitis detection with the thermal infrared images shared the animal welfare, but the detection accuracy needed to be improved during this time. Mastitis detection with the milk images was achieved with high accuracy during lactation SCC detection. But it was still lacking in the test trials on the accuracy of mastitis detection in dairy cows. Secondly, the AMS was often used to collect milk information in the automatic detection of dairy cow mastitis. The local data or manually recorded individual information of dairy cows were then combined to construct the detection model of dairy cow mastitis using machine learning classification. The automation and efficiency of mastitis detection were greatly improved, as well as the accuracy of detection. However, the AMS was easy to cause injury to the dairy cow''s udder during milking, even to the animal welfare and the high cost. Finally, a sensor or multi-sensor system was developed using the rest sensors. A mastitis detection model was constructed using machine learning. The milk or udder data was also obtained to verify, according to the mastitis characteristics and the changes in milk physical and chemical properties. The detection can be expected to fully meet the harsh needs of the rapid and accurate detection of mastitis in dairy cows, due to the low cost and simple operation. Different types of sensors were used to detect mastitis in dairy cows, where the accuracy was quite different. As such, a critical review was proposed on the current progress of accuracy, real-time, and sufficient automation in the detection of mastitis in dairy cows. The future trend was also given to important technical support for future research on the automatic detection of mastitis in dairy cows.