Vision-based Precision Localization of UAVs for Sensor Payload　Placement and Pickup for Field Monitoring Applications

摘要

Due to their mobility and autonomy, unmanned aerial vehicles (UAVs) provide an unprecedented opportunity to performdata gathering in a wide array of civil engineering applications such as visual inspection of infrastructure. Given theirversatility, the role of UAVs can be expanded by leveraging their autonomous operations to deploy wireless sensingresources. This can be especially valuable in numerous field applications such as shear wave velocity (V_s) assessment ofthe subsurface. This study explores the feasibility of automating the autonomous placement and pickup of wirelessgeophone sensors using UAVs for multichannel analysis of surface waves (MASW) for subsurface characterization.Typically, autonomous navigation of UAVs is based on the fusion of inertial sensors and GPS to control the UAV flighttrajectory. However, this approach is not sufficiently accurate for missions requiring precision placement and pickup ofpayloads (such as sensors). Hence, computer vision using fiducial markers can be used to augment traditional inertialsensing to add accuracy to the localization of the UAV relative to payloads. In this study, we use a set of fiducial markersof varying sizes as tracking targets during navigation missions. Pose information extracted from the marker images areintegrated into a sensor fusion controller based on the Kalman filter. The work conducts field validation of the proposedcomputer vision navigation method showing accuracy of the UAV landing on a user defined target within 10 cm; as theUAV descends, smaller fiducial markers are shown to increase the precision of the UAV placement on the ground.