小型无人直升机单目视觉FastSLAM研究  

Research on monocular visual FastSLAM for a small unmanned helicopter

作　　者：王超磊[1] 王田苗[1] 梁建宏[1] 张以成[1] 周易[1] 

机构地区：[1]北京航空航天大学机器人研究所,北京100191

出　　处：《高技术通讯》2013年第10期1061-1067,共7页Chinese High Technology Letters

基　　金：863计划(2011AA040202)资助项目

摘　　要：针对无GPS环境下的小型无人直升机自主飞行问题,提出了基于Rao-Blackwellized粒子滤波(RBPF)的快速同时定位和地图创建(FastSLAM)方法,采用Fast SLAM算法实现了小型无人直升机在GPS环境下的单目视觉SLAM系统。该系统机载的单目摄像头通过尺度不变特征转换(SIFT)算法进行地标的提取和匹配,采用视觉观测和惯导数据融合的方式来对飞机的状态量进行估计同时建立地标地图,采用非延迟的反深度参数化方法来完成地标在三维空间的初始化。仿真实验验证了所用方法的稳定性和有效性。和常规GPS/INS(惯性导航系统)的实际飞行对比实验表明,该系统在姿态、速度和位置等方面均有较高的估计精度,能够在无GPS环境下为小型无人直升机提供可靠的导航信息。For the autonomous flight of a small unmanned helicopter in a GPS-denied environment, a fast simultaneous lo- calization and mapping （ FastSLAM ） algorithm based on Rao-Blackwellized particle filter （RBPF） was designed, and a monocular visual SLAM system for small unmanned helicopters in GPS-denied environments was implemented by using the Fast SLAM algorithm. The onboard monocular camera of the system uses the scale invariant feature trans- form（SIFT） to detect and match landmarks. The visual observation is fused with the inertial measurement to estimate the state of the vehicle and build the feature map simultaneously. An undelayed inverse depth parametrization meth- od is applied to the landmark initialization. The stability and the effectiveness of this system were verified by simula- tions, the real flight experiments were also carried out to test the performance of the algorithm. The results show that the proposed system can estimate the state of the vehicle with higher accuracies in all items such as attitude, velocity and position, compared with the traditional GPS/INS navigation system. It can provide reliable navigation informa- tion for small unmanned helicopters in GPS-denied environments.

关 键 词：小型无人直升机 单目视觉 快速同时定位和地图创建(FastSLAM) 惯性测量单 元(IMU) 

分 类 号：V279[航空宇航科学与技术—飞行器设计]