惯性-视觉-磁场传感器组合的低成本标定方法(英文)

Low cost calibration methods for hybrid inertial-vision-magnetic sensors

针对一种融合了惯性-视觉-磁场信息的传感器组合,给出了一套低成本的标定和建模方法。对于传感器模型中的确定性项和非确定性项参数,分别采用了基于输入指令向量模匹配的方法和基于时间序列分析的方法来进行标定和建模,利用所得到的模型补偿之后的传感器输出误差的方差降为原来的1/3以下;对于摄像机坐标系和加速度计坐标系之间的相对旋转矩阵,提出了一种基于重力向量和铅垂线的标定方法,仿真结果表明,标定结果的相对误差小于7%。

This paper focuses on the low cost calibration methods for a type of hybrid sensors which consists of an MEMS IMU（Inertial Measurement Unit）,a monocular vision sensor and a three-axis magnetic compass.For the calibration of the deterministic items and the modeling of the nondeterministic items,the method of matching the modulus of the input vectors and the method of time series analysis are introduced respectively.The variance of the sensor＇s compensated output errors is reduced to below one third of the original output.For the relative rotation matrix between the accelerometers＇ frame and the camera frame,a calibration method is proposed,which is based on the gravity vector and plumb line.The simulation indicates that the relative errors of the calibration are less than seven percent.