Abstract Automated surveys for wildlife have the potential to improve data collection while averting mortality of animals. Collisions of eagles at wind power facilities are particularly of concern and therefore an automated system that could detect birds, determine if they are eagles, and track their movement, might aid in curtailing wind turbines before collisions occur. Here, we use human observers and photographs to test the ability of a camera-based monitoring system, called IdentiFlight, to detect, classify, and track birds. IdentiFlight detected 96% of the bird flights detected by observers and detected 562% more birds than did observers. The discrepancy between observers and IdentiFlight seemed to be because the ability of observers to detect birds declined sharply by distance and toward the west. We reviewed photographs taken by IdentiFlight and determined that IdentiFlight misclassified nine of 149 eagles as non-eagles for a false negative rate of 6%, and 287 of 1013 non-eagles as eagles for a false positive rate of 28%. The median distance at classification for birds classified as eagles was 793 m and the median time from detection till classification was 0.4 s. Collectively, our results suggest that automated cameras can be effective means of detecting birds in flight and identifying eagles.

中文翻译：

---

自动监测飞行中的鸟类：在风力发电设施中使用老鹰进行概念验证

摘要 对野生动物的自动化调查有可能改善数据收集，同时避免动物死亡。风电设施中鹰的碰撞尤其令人担忧，因此可以检测鸟类、确定它们是否是鹰并跟踪它们的运动的自动化系统可能有助于在碰撞发生之前减少风力涡轮机。在这里，我们使用人类观察者和照片来测试基于摄像头的监控系统（称为 IdentiFlight）检测、分类和跟踪鸟类的能力。IdentiFlight 检测到观察者检测到的鸟类飞行的 96%，检测到的鸟类比观察者多 562%。观察者和 IdentiFlight 之间的差异似乎是因为观察者探测鸟类的能力随着距离和向西而急剧下降。我们查看了IdentiFlight 拍摄的照片，确定IdentiFlight 将149 只老鹰中的9 只误分类为非老鹰，误报率为6%，将1013 只非老鹰中的287 只误分类为老鹰，误报率为28%。被归类为鹰的鸟类在分类时的中位距离为 793 m，从检测到分类的中位时间为 0.4 s。总的来说，我们的结果表明，自动摄像机可以成为检测飞行中的鸟类和识别鹰的有效手段。