【作者】 刘旭； 孙培； 卢山； 苏宏业； 金鑫；

【Author】 LIU Xu;SUN Pei;LU Shan;SU Hongye;JIN Xin;Liaoning Petrochemical University;Shenzhen Polytechnic;Zhejiang University;

【机构】 辽宁石油化工大学； 深圳职业技术学院； 浙江大学；

【摘要】 随着无人机技术的发展,无人机的位姿估计是当下研究的热点。现有基于视觉的位姿估计算法大多利用传统相机所获得的图像信息实现的,但在高速运动的场景中容易产生运动模糊,并且在极端光照条件下图像清晰度差,无法进行有效估计。事件相机具有低延迟,高动态的特性,为此本文提出了一种基于单目事件相机的无人机位姿回溯映射估计方法。本方法将事件流信息进行回溯映射,通过最小化映射后事件坐标方差,获得运动参数的最优估计,并进行坐标转换,从而对高速运动的无人机进行位姿估计。本文最后采用仿真数据验证了算法精确性;并利用动作捕捉系统采集的数据作为真实值对比,验证了对无人机位姿变化的有效跟踪。更多还原

【Abstract】 With the development of UAV technology, the pose estimation of UAV is a hot topic in current research. Most of the existing vision-based pose estimation algorithms are realized by using the image information obtained from traditional cameras. However, motion blur is likely to occur in high-speed moving scenes, and the image clarity is poor under extreme lighting conditions, which cannot be effectively estimated. The event camera has the characteristics of extremely low latency and high dynamics range, so a UAV pose estimation method via warp mapping based on monocular event camera is proposed. The motion parameters can be estimated by warping the events and minimizing their coordinates variance, and the coordinate conversion is performed to obtain the high-speed oving UAV pose. we use the simulation data to verify the accuracy of the algorithm; The data collected by motion capture system is used as the real value comparison, which verifies effective tracking of changes in the position of UAV.更多还原