Fusion of udder temperature and size features for the automatic detection of dairy cow mastitis using deep learning

Abstract

Mastitis seriously reduces the welfare and milk yield of dairy cows and affects the economic value of farms. Most dairy cow mastitis detection algorithms based on computer vision used the threshold classification method to detect mastitis only according to the temperature feature of cows’ udders, and the results were easily affected by the farm environment and individual cow characteristics. In this work, an automatic detection method for dairy cow mastitis using the fusion of udder temperature and size features based on deep learning was proposed. First, you only look once version 7 (YOLOv7) network was used to automatically detect the eye and udder regions of dairy cows, extract the corresponding temperatures, and construct an udder temperature feature vector. Then, the CenterNet network was used to automatically detect the keypoints of the udder region, calculate its corresponding size, and construct an udder size feature vector. Finally, both the temperature and size features of udders were fused, and the support vector machine (SVM) algorithm based on the wrapper was used to automatically detect the degree of mastitis. Thermal infrared videos of 196 dairy cows were used as the dataset for evaluating the performance of the proposed method, and the accuracy of the proposed method was 88.61%. The sensitivity and specificity values of the model for clinical mastitis (CM) were 87.50% and 94.03%, respectively, those for subclinical mastitis (SCM) were 81.25% and 91.94%, respectively. The proposed method constitutes a promising approach for automatic diagnosis in the early stages of dairy cow mastitis, and it can be used for objective monitoring and optimal management of commercial farms.