视觉引导无人机降落系统鲁棒性优化研究  

Application of Visual Guidance and Target Recovery Technology in Autonomous Landing of Unmanned Aerial Vehicles

作　　者：唐嘉泽 国添星 刘丹[1] 王启松[1] 李君宝[1] 邵明媚 孙金玮[1] TANG Jiaze;GUO Tianxing;LIU Dan;WANG Qisong;LI Junbao;SHAO Mingmei;SUN Jinwei(Harbin Institute of Technology,Harbin 150000,China)

机构地区：[1]哈尔滨工业大学,黑龙江哈尔滨150000

出　　处：《探测与控制学报》2024年第6期126-136,共11页Journal of Detection & Control

摘　　要：为解决在复杂环境中无人机在移动平台上自主降落的定位精度问题,设计了一种鲁棒视觉引导靶标恢复的无人机着陆系统(RVGTRDLS)。RVGTRDLS采用四角ArUco技术纠正倾斜的画面,并结合YOLO技术进行降落靶标位置重构。通过卡尔曼滤波技术与降落靶标的定位信息融合,即使在靶标完全丢失的情况下,也能实现对靶标位置的准确再定位。经过多场飞行试验验证,当无人机在降落过程中遭遇靶标信息缺失、靶标框架倾斜、靶标部分或完全丢失的困境时,RVGTRDLS能够通过靶标重构与靶标再定位技术来成功降落。无论是在静态还是动态的着陆场景中,该系统均展现出优异的性能。To address the problem of localization accuracy in realizing autonomous UAV landing on a mobile platform in complex environments,a Robust visually guided target recovery for drone landing system(RGTRDLS)was designed in this paper.It used quadrangle ArUco code technology to correct skewed frames and combined YOLO and SIFT feature point matching techniques for accurate target reconstruction.Through the fusion of Kalman filtering technology and the target s localization information,accurate prediction and tracking of the target could be realized even when the target was completely lost.After multiple flight experiments,the system showed excellent performance in both static and dynamic landing scenarios.Ultimately,the RVGTRDLS had been successfully applied to a real UAV,ensuring its accurate landing on a moving platform.Validated by multiple flight experiments,the system was able to respond effectively when the UAV encountered the challenges of missing target information,tilted target frames,or partial or complete target loss during landing.the system demonstrated excellent performance in both static and dynamic landing scenarios.

关 键 词：卡尔曼滤波 靶标丢失 靶标重建 恢复引导 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]