Body Temperature Detection of Group-Housed Pigs Based on the Pairing of Left and Right Ear Roots in Thermal Images

Body temperature is a critical indicator of pig health. This study proposes a non-contact method for detecting body temperature in group-housed pigs by extracting temperature data from thermal images of ear roots. Thermal images in the drinking trough area were captured using a thermal camera, with real-time data transmitted to a monitoring room via optical fibers. The YOLO v11m-OBB model was utilized to detect the ear root areas with oriented bounding boxes, while a novel algorithm, the two-stage left and right ear root pairing algorithm (YOLO TEPA-OBB), paired the ear roots of individual pigs using center distance clustering and angular relationships in a polar coordinate system. The maximum temperature of the ear roots was extracted to represent the body temperature. Experimental results based on 749 ear roots show that the YOLO TEPA-OBB achieves 98.7% precision, 98.4% recall, and 98.7% mean average precision (<i>mAP</i>) in detecting ear roots, with an ear root pairing accuracy of 98.1%. The Pearson correlation coefficient (<i>r</i>) between predicted and reference temperatures is 0.989, with a mean bias of 0.014 °C and a standard deviation of 0.103 °C. This research facilitates real-time body temperature monitoring and precise health management for group-housed pigs.