AIGLE: Identifying birds species of metropolitan France to better prevent bird collisions at airports

Guillaume Testud, Jean-Philippe Tarel and Sylvain Moulherat

Abstract

---

Collisions between animals and transport have an impact not only on vehicles and their passengers, but also on biodiversity. In the case of airports, collisions between aircraft and birds, and more generally with animals, lead to loss of human life, costs in the order of a billion dollars a year, and mortality among the animals affected. In 2023, more than 1,000 collisions were detected at French airports. While 22% of species are unidentified, the others may include species on the red list of breeding birds in metropolitan France (2016), such as the Northern lapwing Vanellus vanellus (NT) and the Common kestrel Falco tinnunculus (NT), the two species most frequently implicated, or the Little bustard Tetrax tetrax (EN). Furthermore, damage to aircraft is linked to the number and weight of animals impacted. Knowing the species can therefore also help to assess the safety risk associated with collisions. Among the species affected there are some with gregarious behaviour: Common starling Sturnus vulgaris, or species with large weights such as mammals or for exemple Common buzzard Buteo buteo. To reduce the risk of collisions, we propose several image recognition models, trained to detect and identify birds in metropolitan France on YoloV5 from the AIGLE database. Different learning transfer strategies are compared, for example from training on the COCO or megadetector database. The taxonomic hierarchy is used for some models. These models can be connected to existing camera networks in airports. Or to a network of connected sensors communicating with animal hazard management services.

Reference

---