摘要
卫星拒止条件下无人机与机场跑道的相对位姿信息的精确测量是无人机实现重复使用的关键。视觉相对位姿测量以其自主性、精度高的优势,为无人机自主着陆提供有效技术途径。但在实际工程应用中,基于单帧图像提取的跑道视觉特征进行跑道相对位姿测量时,精度及鲁棒性易受环境变化、成图质量等因素影响。针对该问题,提出了一种基于序列帧图像的无人机自主着陆相对位姿测量方法。在单帧图像跑道关键特征提取的基础上,通过连续多帧图像建立序列历史帧与当前帧的单应关系,将序列历史帧图像提取结果映射到当前帧图像上,继而通过优化算法获得稳定准确的跑道特征,最终基于点线综合特征解算无人机相对机场跑道的位置与姿态。机载飞行数据离线仿真结果表明,所提方法前向相对位置测量精度平均提升10%,且在1000~0 m阶段实现米级侧向和垂向相对位置的连续可靠测量,有效提升了无人机着陆相对位姿测量算法的鲁棒性与精度。
The accurate measurement of the relative position and attitude information between the unmanned aerial vehicle(UAV)and the airport runway under satellite denied is crucial to the reuse of the UAV.Visual relative position and attitude measurement,with its advantages of autonomy and high accuracy,provides an effective technical approach for UAV autonomous landing.However,in practical engineering applications,the accuracy and robustness of runway relative position and attitude measurements based on visual features extracted from single-frame image are susceptible to environmental changes,image quality and other factors.To solve this problem,a relative position and attitude measurement method for UAV autonomous landing based on sequential frame images is proposed.Based on the extraction of runway key features from single-frame images,the homography relationship between sequential historical frames and the current frame is established through successive multi-frame images,the extraction results of the sequential historical frame images are mapped to the current frame images,and stable and accurate runway feature extraction results are obtained by optimization algorithm.Finally,the position and attitude of the UAV relative to the airport runway are solved based on point-line synthetic features.The offline simulation results of airborne flight data demonstrate that the proposed method improves the accuracy of forward relative position measurement by 10%on average,and achieves continuous and reliable measurement of meter-level lateral and vertical relative position during the 1000 m to 0 m phase,effectively improving the robustness and accuracy of the UAV landing relative position and attitude measurement algorithm.
作者
尚克军
魏永树
扈光锋
徐策
明丽
SHANG Kejun;WEI Yongshu;HU Guangfeng;XU Ce;MING Li(Beijing Institute of Automation and Control Equipment,Beijing 100074,China)
出处
《中国惯性技术学报》
北大核心
2025年第5期455-461,共7页
Journal of Chinese Inertial Technology
基金
国防科技重点实验室基金项目(2021-JCJQ-LB-070-16)。
关键词
无人机视觉着陆
序列帧
单应关系
相对位姿测量
UAV visual landing
sequential frame
homography
relative position and attitude measurement
