基于直线稀疏光流场的微小型无人机姿态信息估计方法研究

针对无人飞行器视觉导航系统的需求，提出一种基于直线稀疏光流场的微小型无人机姿态信息估计方法。研究了图像直线稀疏光流场的概念，并推导出该种稀疏光流场的计算方法；在直线稀疏光流场的基础上，建立了地平线投影模型，并提出一种基于直线稀疏光流场的飞行器俯仰角速度、滚转角速度和偏航角速度等姿态信息的估计方法；分别使用文中提出的直线稀疏光流场的计算方法和Horn算法对一组测试图像进行数值仿真，仿真结果显示在图像边缘附近，直线稀疏光流场的计算方法与经典的Horn算法具有相近的精度，但时间开销减小到后者的6%；使用无人机航拍图像序列对于基于直线稀疏光流场的飞行器姿态信息算法进行离线仿真，并对比同时搭载的角速度陀螺的测量结果，结果显示，使用文中所述的基于直线稀疏光流场的无人机姿态提取方法可以估计飞行器的俯仰、偏航、滚转三通道角速度信息，其最大绝对估计误差除滚转通道小于±10°/s外， 其余两通道均小于±5°/s，并对于误差产生的原因进行了进一步讨论。

Research on Attitude Estimation of Micro UAV Based on Sparse Line Optical Flow Field

A method to estimate the attitude of micro UAV based on the sparse line optical flow field is proposed according to the requirements for UAV vision-based navigation system.The concept of sparse line optical flow field of images is described, and an algorithm of sparse line optical flow field is presented. A new method to estimate the attitudes of micro UAV, including pitch rate, roll rate and yaw rate, is proposed based on the sparse line-optical flow field and the established horizon projection model. The proposed method and the classic Horn algorithm are used to numerically simulate a group of test images for calculating the optical flow field. The results show that the proposed method has the same calculation accuracy as the classic Horn algorithm, while the calculating time-cost of the former is only 6% of Horn algorithm. Taking an aerial image sequence as test samples, an off-line simulation is conducted to verify the method of UAV attitude estimation based on sparse line-optical flow field. The information calculated with the measured values of angular velocity gyro on UAV is compared. The comparative result shows that the proposed method is efficient for estimation of pitch, roll, and yaw rates of UAV. The calculated error of pitch rate is less than ±10(°), while the errors of roll and yaw rates are less than ±5(°)/s .