无人机视觉SLAM算法及仿真  

UAV vision SLAM algorithm and simulation

作　　者：王希彬[1] 赵国荣[1] 寇昆湖[1] 

机构地区：[1]海军航空工程学院控制工程系,山东烟台264001

出　　处：《红外与激光工程》2012年第6期1653-1658,共6页Infrared and Laser Engineering

基　　金：国防预研基金(51309060401)

摘　　要：为将SLAM算法从地面机器人的两维空间扩展到无人机的三维环境,研究了SLAM的发展现状,指出无人机视觉SLAM面临的挑战,利用摄像机和惯导传感器的组合,建立了无人机视觉SLAM算法的数学模型;并提出了一种改进的延迟地标初始化的方法,为了便于理解和实施,简化了系统的状态和方差阵的结构,减少了存储和计算量。用EKF滤波对无人机视觉SLAM进行了仿真研究,得到了较好的估计效果,表明了SLAM在无人机上应用的可行性。In order to expand SLAM algorithm from 2D space of ground robot to 3D environment of uninhabited aerial vehicle （UAV）, the latest progress on simultaneous localization and mapping was researched, the challenge that UAV vision SLAM faces up with was pointed out, and the mathematic model of UAV vision SLAM algorithm was built with the integration of camera and IMU. Furthermore, an improved delayed landmark initialization method was proposed which was easily understood and implemented, thus the system structure of state and variance matrix was simplified, and the storage and computation quality were reduced. Simulation research was carried out by extended Kalman filter on UAV vision SLAM, and better estimation effect was gotten, demonstrating the feasibility of SLAM on UAV.

关 键 词：同时定位与作图 无人机 视觉 地标初始化 

分 类 号：V24[航空宇航科学与技术—飞行器设计]