孙楠, 裴信彪, 王春军, 李继辉, 彭程, 白越. 基于立体视觉—惯导SLAM的四旋翼无人机导航算法[J]. 微电子学与计算机, 2020, 37(5): 33-38,42.
引用本文: 孙楠, 裴信彪, 王春军, 李继辉, 彭程, 白越. 基于立体视觉—惯导SLAM的四旋翼无人机导航算法[J]. 微电子学与计算机, 2020, 37(5): 33-38,42.
SUN Nan, PEI Xin-biao, WANG Chun-jin, LI Ji-hui, PENG Cheng, BAI Yue. Four-rotor UAV navigation algorithms based on stereo visual-intertial SLAM[J]. Microelectronics & Computer, 2020, 37(5): 33-38,42.
Citation: SUN Nan, PEI Xin-biao, WANG Chun-jin, LI Ji-hui, PENG Cheng, BAI Yue. Four-rotor UAV navigation algorithms based on stereo visual-intertial SLAM[J]. Microelectronics & Computer, 2020, 37(5): 33-38,42.

基于立体视觉—惯导SLAM的四旋翼无人机导航算法

Four-rotor UAV navigation algorithms based on stereo visual-intertial SLAM

  • 摘要: 针对基于视觉的同时定位与地图构建(SLAM)系统在快速运动和特征不足时存在精度低和跟踪易失败的问题,提出了一种基于紧耦合的非线性优化的立体视觉—惯导SLAM(visual inertial SLAM, VI-SLAM)算法.首先,以关键帧的位姿作为约束,初始化惯性测量单元(IMU)的偏差,并根据IMU预积分预测当前帧的位姿;然后,在后端优化中采取非线性局部平滑的方法,融合视觉SLAM的位姿估计与IMU预积分;并通过回环检测消除累积误差,采用图优化算法进行全局地图优化.通过EuRoC数据集验证该系统的性能,对比视觉SLAM算法平均精度提升了1倍左右.并将该算法应用于立体视觉四旋翼飞行控制平台,验证所提算法的实际应用性.

     

    Abstract: In the case of too fast movement and lack of features, the simultaneous localization and mapping (SLAM) system based on visual features has low accuracy and is liable to fail in tracking.For this problem, a tightly coupled stereo visual-inertial SLAM (VI-SLAM) algorithm using nonlinear optimization is proposed.Firstly, with the pose of the key frame as the constraint, the bias of the Inertial Measurement Unit (IMU) is initialized.and the pose of the current frame is predicted based on the IMU pre-integration.then, the non-linear local smoothing method is used in the back end optimization, fusing the pose estimation of vision SLAM and IMU pre-integration.and the cumulative error is eliminated by loop detection, the global map optimization is carried out by graph optimization.The performance of the system is validated by EuRoC dataset, and the average accuracy of the system is about twice as high as that of the visual SLAM algorithm.The algorithm is applied to the stereo vision Four-rotor UAV platform, and the practical application of the proposed algorithm is verified.

     

/

返回文章
返回