首页|期刊导航|控制理论与应用|低空安全下的无人机感知控制一体化跟踪技术

低空安全下的无人机感知控制一体化跟踪技术OA

Integrated perception and control for UAV target tracking in low-altitude airspace security

中文摘要英文摘要

低慢小无人机具有尺度小、机动性强、低空背景复杂等特点,易造成视觉漏检、误检和跟踪中断,影响低空安全防御任务的连续性.针对复杂城市与关键基础设施背景下低慢小目标难以实时检测和稳定跟踪的问题,提出一种无人机视觉感知与跟踪控制一体化方法.首先,基于YOLOv10构建轻量化检测网络BiGNet,通过多尺度特征融合和轻量化卷积增强复杂背景下小目标表征能力,并降低模型计算开销.然后,将视觉检测结果经坐标变换转换为目标相对位置信息,并在模型预测控制框架中构造视觉感知代价项,使目标图像中心偏差和期望跟踪距离误差参与滚动优化.进一步地,针对近距离跟踪过程中的气动扰动影响,采用高斯过程回归建模附加扰动,并将扰动补偿引入预测模型.仿真与对比实验表明,所提方法能够提高复杂低空场景下低慢小目标的检测效果和在线推理效率,并在扰动条件下保持目标可见性和跟踪稳定性,为低空安全防御中的持续跟踪与后续处置提供技术支撑.

Low-slow-small unmanned aerial vehicles are difficult to detect and track in low-altitude airspace because of their small visual scale,agile motion,weak texture features,and complex backgrounds.These characteristics may lead to missed detections,false alarms,and tracking interruptions,thereby affecting the continuity of low-altitude security op-erations.To address real-time detection and stable tracking of such targets in urban and critical-infrastructure scenarios,an integrated visual perception and tracking control method is proposed.Firstly,a lightweight detection network named BiGNet is developed based on YOLOv10,where multi-scale feature fusion and lightweight convolution are introduced to improve small-target representation while reducing computational cost.Then,the visual detection results are transformed into relative target position information,and a visual perception cost is constructed in a model predictive control frame-work.In this way,image-center deviation and desired tracking-distance error are included in receding-horizon optimization.Furthermore,Gaussian process regression is used to model aerodynamic disturbances during close-range tracking,and dis-turbance compensation is incorporated into the prediction model.Simulation and comparison results show that the proposed method improves detection and online inference performance in complex low-altitude scenarios,while maintaining target visibility and tracking stability under disturbances.

陈子路;梁鸿雅;张毅博;韩亮;李东禹;胡庆雷

北京航空航天大学 网络空间安全学院,北京 100191北京航空航天大学 中法工程师学院,北京 100191北京航空航天大学 中法工程师学院,北京 100191北京航空航天大学 中法工程师学院,北京 100191北京航空航天大学 网络空间安全学院,北京 100191北京航空航天大学 自动化科学与电气工程学院,北京 100191

低慢小目标无人机目标跟踪模型预测控制目标检测高斯过程回归

low-slow-small targetsunmanned aerial vehicles(UAV)target trackingmodel predictive controlobject detectionGaussian process regression

《控制理论与应用》 2026 (9)

1868-1880,13

国家自然科学基金项目(62373024)资助. Supported by the National Natural Science Foundation of China(62373024).

10.7641/CTA.2026.60037

评论