基于惯性光流的多旋翼飞行器地形跟踪

Terrain-following for multi-rotor aircraft based on inertial optical flow

为解决目前多旋翼飞行器光流技术存在的定位精度低与多相机数据处理过程繁琐等问题,本文应用金字塔层数自适应的Lucas-Kanade算法获取平移惯性光流以实现多旋翼飞行器稳定地形跟踪。下视球面相机获取多纹理的目标平面图像,同时将目标平面特征点动态映射至相机镜头表面;将比对两帧连续时刻图像后获取的平移光流作为非线性地形跟踪控制器的反馈,实现多旋翼飞行器地形跟踪与避免碰撞。10次室内实际飞行测试结果表明,不同地形变化区域的高度误差分别为±3 cm与±6 cm,验证了该控制器是稳定可行的。

To solve the optical flow technique problems of low positioning accuracy and cumbersome processing of multi-camera data in multi-rotor aircraft for the moment,the layer adaptive pyramid Lucas-Kanade algorithm is applied to get the translational inertial optical flow to achieve the stability terrain-following of multi rotor aircraft in this paper. The image of the textured target plane was captured by the bottom facing spherical camera. Simultaneously,the feature points in the target plane were dynamically mapped to the camera lens surface. Then the translational optical flow obtained from the comparison between the two time continuous images was used as feedback of the nonlinear terrain-following controller to achieve the terrain-following and collision avoidance for multi-rotor aircraft. The result of 10 indoor actual flight tests showed that the height accuracies in different changing terrains are ±3 cm and ±6 cm respectively. And they verified that the nonlinear controller proposed in this paper is stable and feasible.