Abstract :
Time-of-Flight (ToF) cameras provide 3D point-cloud data in real time, even in case of changing illumination conditions. Thanks to their advantages in comparison to classic vision techniques, ToF devices have been already used in several application fields but their suitability in aerial scenarios is still under study. Thus, this paper presents the evaluation of a new approach for detecting an aerial target within the working range of the used ToF camera, defining a boom-and-receptacle refueling system as the context of this work. The detection method is based on a-contrario reasoning, using the 3D point-cloud data acquired by a PMD CamCube 3.0: the sky is presented as noise in the range images, thus the target is perceived as a deviation from a model of complete randomness. In order to obtain a decision criterion, a study based on mean and variance analysis has been carried out. The results on real images show that the detection is performed successfully with the proposed approach, showing its robustness with respect to the number of false alarms.
