用于无人机室内导航的光流与地标融合方法
Optical flow and landmark fusion method for unmanned aerial vehicles indoor navigation

针对小型无人机在无卫星导航信号条件下的导航问题, 结合光流及地标定位设计了使用摄像头、惯性测量 器件、超声测距仪等传感器融合的无人机室内导航方法. 文章使用补偿角速率的光流微分法计算帧间像素点小位 移, 并用前后误差算法提取精度较高的点, 避免像素点跟踪错误, 提高了光流测速的精度; 对得到的光流场用均值漂 移算法进行寻优, 得到光流场直方图峰值, 以此计算光流速度. 本文提出了无累积误差的连续地标定位算法, 实时测 量无人机位置. 通过多速率卡尔曼滤波器对观测周期不一致的位置、速度信息进行最优估计. 在搭建的八旋翼无人 机平台上试验, 将位置与速度测量结果分别与激光和PX4FLOW数据对比, 结果表明该导航方法可以有效抑制定位 跳变与光流测量噪声误差, 给出精确的位置与速度估计.
Combining optical flow with landmark localization, we propose an indoor navigation methodology fusing camera, inertial measurement unit (IMU) and ultrasonic range finder, for the navigation of unmanned aerial vehicles (UAV) in global navigation satellite systems (GNSS)-denied environment. We employ the differential optical flow algorithm compensated with the angular velocity to determine the small shift between the frames, and use the forward-backward error algorithm to detect the reliable trajectories and avoid the tracking failure, which improves the precision of the optical flow velocimetry. Optimizing the obtained optical field by using the mean shift algorithm, we determine the peak values of the flow field histogram, from which we can calculate the optical flow velocity. Furthermore, we propose in this research the consecutive landmark-localization algorithm with no error accumulation to measure the position of the UAV in real time. By means of the multi-rate Kalman filter, we optimally estimate the information of UAV position and velocity with un-uniform observation periods. Comparing the experiment results with the laser and PX4FLOW data on multi-rotor UAV platform, we find that the proposed methodology can suppress the localization hopping and reject the measurement noise of optical flow, and provide precise estimation of position and velocity.