无人机软管式自主空中加油视觉导航技术

Visual Navigation Technology for UAV Autonomous Hose-Drogue Aerial Refueling

关键词:  无人机；自主空中加油；导航；加油锥套  位姿

Key words:  UAV  Autonomous Aerial Refueling  Navigation  Drogue  3D-pose

摘要:

近年来，机器视觉技术的迅速发展给无人机自主空中加油提供了一种新思路，克服了传统GPS动态适应能力差的不足。基于双目视觉，提出了一种实用的无人机自主空中加油导航方案，并提出了一种利用深度学习进行检测跟踪配合几何约束进行位姿计算的技术。根据加油机和受油机距离的不同，将导航分为远端导航和近端导航。全程对加油锥套进行检测和跟踪，并在近端进行锥套位姿计算。实验结果表明，该算法满足加油的鲁棒性要求，稳定跟踪锁定目标距离范围大于40m。当与目标距离小于10m时，距离误差优于±10cm。说明本文提出的方法具有较强的应用前景。

Abstract:

In recent years, the rapid development of machine vision technology provides a new idea for autonomous aerial refueling of UAV, which overcomes the shortcomings of the poor dynamic adaptability of GPS. A practical technical solution based on binocular vision for UAV autonomous aerial refueling navigation is proposed. Based on binocular vision, a technique of detecting, tracking based on deep learning and pose calculation based on geometric constraint is proposed. The navigation is divided into remote navigation and near-end navigation according to the distance between the tanker and the receiver. The drogue is detected or tracked all the time and the pose of drogue is calculated at the at the proximal. The experimental results show that the algorithm satisfies the robustness requirements of fueling and the range of stable tracking target distance is more than 40 m. When the distance between the receiver and the target is less than 10 m, the distance error is better than ±10 cm. It shows that the method proposed in this paper has a strong application prospect.