Abstract
Automated animal monitoring is a growing segment in animal agriculture. These technologies capture biometric data from numerous parameters such as facial recognition, rumination, body temperature, and animal movement. The goal of automated animal monitoring technologies is to provide a 24 h perspective on an individual animal. These devices can offer valuable information pertaining to animal health, location, locomotion, animal behavior, phenotypic traits, and reproductive cycles. Devices collect data in various forms, such as counting steps, G.P.S. coordinates, triangulation, or XYZ coordinates. Raw data (e.g. step counts, XYZ coordinates) are then analyzed with proprietary algorithms to provide information in a usable format to the consumer. However, the development of these algorithms require thousands of hours of observation. For example, thorough visual observation of individual animal behaviors (e.g. rumination and eating time) can be identified based upon accelerometer movement. However, these calculated behaviors must then be validated through observation to optimize accuracy. Accelerometer-based devices are common in the dairy industry and growing in popularity in the beef industry for monitoring health and detecting estrus. Twenty-four-hour observation provides livestock producers greater insight than does once or twice daily observations. However, the usefulness or reliability of a technology requires evaluation of outcomes based on sensitivity and specificity of the test. The choice to optimize sensitivity or specificity is dependent on the desired outcome of the end user. For example, by increasing sensitivity in an animal health monitoring system, more truly ill animals will be identified, but more false positives will be generated, resulting in more animal treatments and elevated treatments costs. Conversely, increasing specificity, sick animals will be missed, but treatment costs will be reduced. Ultimately, developing an automated animal monitoring system that meets consumer expectation is the goal; however, information interpretation and evaluation will still be necessary by the end user.