基于EKF的摄像机-IMU相对姿态标定方法

EKF-BASED RELATIVE POSE CALIBRATION FOR CAMERA AND IMU

摄像机与惯性传感器之间的相对姿态标定是视觉-惯性混合跟踪器的关键技术之一,是混合跟踪器进行数据融合获得鲁棒姿态输出的前提。提出一种新颖的基于扩展卡尔曼滤波器EKF(Extended Kalman Filter)的摄像机-惯性测量单元IMU(Internal Measurement Unit)相对姿态标定方法。该方法通过构建基于刚体运动学的过程模型和基于摄像机外参数的测量模型,估计摄像机与惯性传感器的相对位置和方向。初步实验结果显示,所提出的标定方法不仅能够标定6 DOF相对姿态,标定操作更简易快速,而且在系统初始误差较大和非线性噪声较大的条件下,该方法仍然能够精确地获得摄像机与IMU之间的相对姿态。

Calibration of relative pose between camera and inertial sensors is one of the key techniques of vision-inertia hyper tracker, and is also the precondition of hyper tracker getting robust pose output through data fusion. Based on the extended Karman filter （ EKF）, we introduce a novel method of camera-IMU relative pose calibration method. It estimates the relative position and direction of camera and inertial sensors by constructing a transition model based on rigid body motion theory and a measurement model based on camera external parameters. Primary experimental results show that the proposed calibration method can calibrate 6 DOF relative pose by more simple and quick calibration operation, besides, it can also precisely get the relative pose between camera and IMU even there are large initial systematic errors and serious nonlinear noises.