Prediction of microunmanned aerial vehicle flight behavior from two-dimensional intensity images

Abstract

The increasing number of microunmanned aerial vehicles (MUAVs) is a rising risk for personal privacy and security of sensitive areas. Owing to the highly agile maneuverability and small cross section of the MUAV, effective countermeasures (CMs) are hard to deploy, especially when a certain temporal delay occurs between the localization and the CM effect. Here, a reliable prediction of the MUAV flight behavior can increase the effectiveness of CMs. We propose a pose estimation approach to derive the three-dimensional (3-D) flight path from a stream of two-dimensional intensity images. The pose estimation in a single image results in an estimation of the current position and orientation of the quadcopter in 3-D space. Combined with flight behavior model, this information is used to reconstruct the flight path and to predict the flight behavior of the MUAV. In our laboratory experiments, we obtained a standard deviation between 1 and 24 cm in a five-frame prediction of the 3-D position, depending in the actual flight behavior.