基于IMU与单目视觉融合的姿态测量方法

Hybrid Pose Measurement Based on Fusion of IMU and Monocular Vision

快速准确测量运动目标的姿态,在航天航空、机器人等领域应用广泛.针对惯性姿态测量速度快但精度不足而视觉姿态测量精度高但速度慢的特点,提出了一种基于IMU与单目视觉融合的姿态测量方法.IMU姿态测量时,使用欧拉角迭代公式解算运动目标的姿态;单目视觉姿态测量时,使用POSIT算法求解被测目标的姿态;再使用H?滤波器融合两者的测量结果;用融合结果与惯性测量结果的差值修正并更新漂移误差曲线.根据被测目标的两组匀速转动信息,提出了双矢量正交化标定法标定IMU坐标系到目标坐标系的旋转矩阵;根据单目视觉采集的3幅图像信息,提出了三图快速标定法标定靶标坐标系到目标坐标系的旋转矩阵.实验表明,提出的姿态测量方法能实现快速姿态测量,测量精度高.

Quick and accurate object-pose measurement is widely used in aerospace and robot fields.Inertial pose measurement,using Euler angle iteration formula,has an advantage in measuring velocity but suffers from slow drift.In contrast,visual pose measurement based on POSIT algorithm is slow but accurate.This paper proposes a method for measuring attitude by integrating the data from the two types of measurements. H∞ filter is used for fusion.In this paper,Euler drift error curve,caused by the drift of gyroscopic sensor,is corrected and updated according to the difference between the hybrid measurement result and the output of inertial measurement.In order to calibrate the coordinate relationships in pose measurement system,this paper promotes a double-vector orthogonal calibration method and a three-picture fast calibration method.Those methods use the output of IMU and the information from three captured pictures when the object takes a determined rotation.Experimental results show that hybrid pose measurement with fusion resolution is quicker than visual orientation measurement,with higher accuracy.