Flight trajectories of the Asian hornet and honey bees
The RGB-D videos collected by the stereo-cameras were processed using the method presented in Chiron et al. (2013) resulting in a dataset made of the trajectories of every insect flying in front of the beehive. The subject of each trajectory (either Asian hornet or honey bee) was determined using both the flight dynamics such as max speed (i.e. using a clustering approach inspired by Chiron et al. 2014) as well as appearance features such as the body size. More details on size/depth ratio of targets as observed on RGB-D images are available in Requier et al. (2016). Finally, for each trajectory, we extracted the period when it occurred (date, start time), the flight speed (m.s-1), the orientation (3D vector), the curvature (m), and the proportion of time spent in static flight (hovering). The Asian hornet count for each measurement was made by counting them in the five seconds before and after the measurement point. We differentiated hovering behaviour into two categories: those where the individual remained very steady (threshold 1 = less than 2 mm of drifting between 2 images), and those where the individuals drifted slightly from their initial position (threshold 2 = less than 10mm of drifting between 2 images).