论文部分内容阅读
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。
In order to solve the problems of low positioning accuracy and complicated data processing of multi-camera data, the Lucas-Kanade algorithm with pyramid level adaptation is used to obtain the translational inertial optical flow so as to realize the stable terrain tracking of multi-rotor aircrafts. The lower spherical camera acquires the multi-textured target plane image and dynamically maps the target plane feature points to the camera lens surface. The translational optical flow obtained after comparing the images of two consecutive frames is used as the feedback of the nonlinear terrain following controller Multi - rotor aircraft terrain tracking and collision avoidance. The test results of 10 indoor real-flight tests show that the height errors of different terrain changes are ± 3 cm and ± 6 cm, respectively, which verifies that this controller is stable and feasible.