无人机动平台着陆惯性/视觉位姿歧义校正算法

Aiming at the application scenario of UAV mobile platform landing and recycling, a visual/inertial navigation solution based on fiducial markers is proposed. The solution focus on solving problems such as the rapid positioning and recognition of markers under complex background conditions, so as to improve the recovery and navigation accuracy, robustness and practicability of UAV platform. First, multiple fiducial markers are designed to construct non-coplanar fiducial marker sequences, and the non-coplanar 3D-2D point pairs are used to break the coplanar constraint of the pose feature points. Then, infinitesimal plane-based pose estimation algorithm (IPPE) is used to estimate fiducial marker plane pose. Finally, the attitude constraint is introduced by fusing inertial information to solve the ambiguity problem of relative pose estimation during the UAV landing and recovery process. Multiple flight experiments show that the proposed navigation scheme can effectively filter out complex background interference, repair the rope occlusion, eliminate pose ambiguity and has good robustness. The operating distance is greater than 30 m, the recovery end reaches the positioning accuracy of centimeter-level, which meets the needs of unmanned aerial vehicle moving platform landing and recovery, and has good application prospects.