Center for Animal Welfare, Department of Animal Science, University of California, Davis 95616.
Department of Computer Science, University of California, Davis 95616.
J Dairy Sci. 2017 Oct;100(10):8430-8437. doi: 10.3168/jds.2017-12984. Epub 2017 Aug 10.
Healthy cattle readily use grooming brushes but this behavior subsides when animals become ill. Tracking use of a brush may provide an opportunity for health monitoring, especially if the process could be automated. We assessed how healthy heifers groom themselves on a brush and hypothesized that radiofrequency identification (RFID) could be used to accurately and automatically record this behavior. Angus and Hereford heifers (n = 16) were fitted with 2 ultra-high-frequency RFID ear tags and monitored in groups of 8 while housed in a pen with an electronic brush, video cameras, and 4 RFID antennas. Each heifer was observed for a 6-h period using continuous video recordings, and brush contact was characterized in terms of anatomic region involved (head/neck, trunk, or posterior) and when not touching the brush but within 1 body length of it. The RFID data were collected for the same period and then processed such that intervals of up to 16 s with no detections but contained between 2 recordings were also considered positive (animal in brush proximity). Brush proximity (RFID) was regressed against brush contact duration (video) and the sensitivity and specificity for each individual heifer calculated. Across heifers, the majority of brush use involved the head/neck, although a few heifers demonstrated relatively large amounts of posterior contact, which contributed to false-negative readings when antennas failed to read the ear tags. Furthermore, for the majority of time that animals were near the brush, they were not in contact with it but rather standing or lying nearby, resulting in false-positive readings. It follows that the ability of the RFID system to accurately detect brush contact varied widely across individual heifers (sensitivity 0.54-1.0; specificity 0.59-0.98), with RFID generally overestimating the duration of brush proximity relative to actual time spent in brush contact. The implication is that RFID-based ear tag recording of brush proximity relative to continuous video observations of contact does not yield accurate results in certain heifers and therefore, as currently configured, is not a reliable representation of this type of grooming behavior.
健康的牛很容易使用梳理刷，但当动物生病时，这种行为就会减少。跟踪梳理刷的使用情况可能为健康监测提供机会，特别是如果这个过程可以自动化的话。我们评估了健康小母牛在梳理刷上的自我梳理情况，并假设射频识别 (RFID) 可用于准确和自动记录这种行为。安格斯和海福特小母牛（n = 16）佩戴了 2 个超高频 RFID 耳标，并在一个带有电子刷、摄像机和 4 个 RFID 天线的围栏中以 8 头为一组进行监测。使用连续视频记录对每头小母牛进行了 6 小时的观察，根据所涉及的解剖区域（头部/颈部、躯干或后部）以及未接触刷子但在 1 个体长内的情况，对刷子接触进行了描述。同时收集了相同时间段的 RFID 数据，并对其进行了处理，即如果没有检测到 16 秒的间隔，但在两次记录之间，也将其视为阳性（动物接近刷子）。将刷子接近度（RFID）与刷子接触持续时间（视频）进行回归，并计算每头小母牛的灵敏度和特异性。在所有小母牛中，大多数刷子使用都涉及头部/颈部，但有少数小母牛表现出相对较大的后部接触，这导致当天线无法读取耳标时出现假阴性读数。此外，在动物靠近刷子的大部分时间里，它们并没有与刷子接触，而是站在或躺在附近，导致假阳性读数。因此，RFID 系统准确检测刷子接触的能力在个体小母牛之间差异很大（灵敏度 0.54-1.0；特异性 0.59-0.98），与实际刷子接触时间相比，RFID 通常会高估刷子接近度的持续时间。这意味着，基于 RFID 的耳标记录相对于连续视频观察到的接触，在某些小母牛中无法产生准确的结果，因此，按照当前的配置，它不是这种梳理行为的可靠代表。
