视觉和惯导信息融合小型无人机位姿估计研究

Pose estimation of small UAV based on vision and INS information fusion

针对无人机在卫星信号失锁时仅利用纯惯导无法精确获取位姿信息的问题,提出通过扩展卡尔曼滤波对单目视觉和惯导进行信息融合的算法,并设计一个包含单目视觉、IMU及超声波传感器的组合系统。首先,利用超声波传感器通过最小二乘法完成尺度估计值的获取;其次建立IMU的系统方程进行惯导信息的解算,并完成误差状态方程的求解;最后,通过扩展卡尔曼滤波实现单目视觉与惯导的信息融合。利用该算法对无人机在200 m动态飞行的信息进行解算,实验结果表明,位置误差的均方根在0.995 m以下,水平姿态角的均方根误差在1.915°以下,偏航角的均方根误差达到了2.235°。不但提高了位姿的估计精度,而且解决了纯视觉输出频率低的问题,满足了无人机对高动态特性的需求,可解决无人机在卫星信号失锁时无法精确定位的问题。

In order to solve the problem that the position and attitude information of UAV can not be accurately obtained only by using pure inertial navigation when the satellite signal is out of lock,an information fusion algorithm based on extended Kalman filter for monocular vision and inertial navigation is proposed,and an integrated system including monocular vision,IMU and ultrasonic sensor is designed.Firstly,the ultrasonic sensor is used to obtain the scale estimation by the least square method.Secondly,the IMU system equation is established to calculate the ins information,and the error state equation is solved.Finally,the information fusion of monocular vision and INS is realized by extended Kalman filter.The experimental results show that the root mean square error of position error is less than 0.995 m,and the root mean square error of horizontal attitude angle is 1.915°.The root mean square error of yaw angle is 2.235°.It not only improves the accuracy of pose estimation,but also solves the problem of low frequency of pure vision output,meets the needs of UAV for high dynamic characteristics,and solves the problem that UAV can not accurately locate when the satellite signal is out of lock.