Kim Sueun, Yamagishi Norio, Ishikawa Shingo, Tsuchiaka Shinobu
Laboratory of Large Animal Clinical Medicine, Graduate School of Veterinary Sciences, Osaka Metropolitan University, Osaka, Japan.
Front Vet Sci. 2025 Mar 10;12:1548906. doi: 10.3389/fvets.2025.1548906. eCollection 2025.
This study investigates the potential of non-invasive, continuous temperature measurement techniques for assessing cattle welfare. We employed advanced object detection algorithms and infrared thermography to accurately extract and continuously measure temperatures of the eyes and muzzles of 11 calves over several months (total, 33 samples). A mobile thermal imaging camera was paired with the Mask R-CNN algorithm (object detection) trained on annotated datasets to detect eye and muzzle regions accurately. Temperature data were processed by outlier rejection, standardization, and low-pass filtering to derive temperature change patterns. Cosine similarity metrics and permutation tests were employed to evaluate the uniqueness of these patterns among the individuals. The average cosine similarity between eye and muzzle temperature changes in the same individual across 33 samples was 0.72, with permutation tests yielding -values <0.01 for most samples, indicating pattern uniqueness. This study highlights the potential of high-frequency, non-invasive temperature measurements for detecting subtle physiological changes in animals without causing distress.
本研究调查了用于评估牛福利的非侵入性连续温度测量技术的潜力。我们采用先进的目标检测算法和红外热成像技术，在几个月的时间里准确提取并连续测量了11头小牛眼睛和口鼻的温度（总共33个样本）。一台移动热成像相机与在注释数据集上训练的Mask R-CNN算法（目标检测）配对，以准确检测眼睛和口鼻区域。温度数据通过异常值剔除、标准化和低通滤波进行处理，以得出温度变化模式。采用余弦相似性度量和排列检验来评估这些模式在个体之间的独特性。在33个样本中，同一个体眼睛和口鼻温度变化之间的平均余弦相似性为0.72，排列检验对大多数样本产生的值<0.01，表明模式具有独特性。本研究强调了高频非侵入性温度测量在检测动物细微生理变化而不造成痛苦方面的潜力。
