基于惯性/激光测距/视觉里程计组合的高空场景尺度误差估计方法

惯性/视觉感知信息融合导航定位技术是目前实现无人机不依赖卫星自主导航的最有效手段。但对于面向高空场景的大型无人机，惯性器件误差与视觉里程计尺度误差耦合且特征平面化导致可观测性下降。针对这一问题，提出了利用惯性/激光测距/视觉里程计组合实现尺度误差估计的方法。通过开展误差模型建立、激光测量点与图像中位置匹配、无人机平飞机动下系统可观测性分析等关键技术研究，实现了高空场景下尺度误差的精确估计。经过300m高度机载试验数据验证，算法精度优于1.5%D，对卫星拒止条件下高空无人机自主导航具有重要意义。

A High-Altitude Scene Scale Error Estimation Method Based on Inertial/Laser Ranging/Visual Odometry Combination

Inertial/visual perception information fusion navigation and positioning technology is currently the most effective means to achieve autonomous navigation of unmanned aerial vehicles (UAVs) without reliance on satellites. However, for large UAVs facing high-altitude scenes, the inertial device error is coupled with the scale error of the visual odometry, and the feature planarization leads to a decrease in observability. Aiming at this problem, a method of using inertial/laser ranging/visual odometry combination to achieve scale error estimation is proposed. Researches on the key technologies such as error model establishment, matching of laser measurement points and positions in images, and system observability analysis under UAV horizontal plane movements are carried out, and accurate estimation of scale errors in high-altitude scenes is achieved. Verified by airborne experiment data at 300m altitude, the accuracy of the algorithm is better than 1.5%D, which is of great significance to the autonomous navigation of high-altitude UAVs under the condition of satellite rejection.