Research On Automatic Calibration Of Realistic Camera - IM Relative Attitude

Augmented reality technology is a new human-machine interaction technology developed on the basis of virtual reality technology. It can enhance the user experience by estimating the observer’s location in a3D scene, adding virtual3D components, images, text or other information generated in real time. The fundamental part of augmented reality technology implementation is able to obtain real-time camera position in space. But it is difficult to obtain stable and accurate camera position information through a single camera visual tracking method. Through the strictly binding combination of a camera and an inertial measurement unit (IMU) and the data fusion of these two types of sensors, we can significantly enhance the stability and accuracy of camera pose tracking operation. Data measured by different sensors usually locates in different coordinate systems. In order to fuse these different type of data, we need to convert different type of sensor data into a unified coordinate system. The coordinate system conversion is basic on known relative position and orientation between the different coordinate systems. Thus, the accuracy of the relative pose of different sensors has a direct impact on the accuracy of the hybrid sensor tracking application.In this paper, an automatic relative pose calibration method between camera and IMU based on extended Kalman filter (EKF) is introduced. By constantly estimating the position and orientation of the IMU in the world coordinate system, the camera pose in IMU coordinate system, IMU velocity in the world coordinate system and the IMU drift error of the linear acceleration and angular velocity in EKF, we can obtain a final calibration results. Simulation and Experimental results show that this method can calibrate6degrees of freedom (DOF) relative pose information between camera and IMU and the calibration operation can be easily and quickly actualized. It also can get valid and precise results even there are large initial systematic errors or serous non-linear noises.The main contents of this paper include:1. Internal and external camera parameters calibration and feature points matching. Achieved by redesigning the camera calibration chessboard can automatically identify the origin of the world coordinate system and the direction of X-axis and Y-axis.2. Design the prediction model of camera-IMU relative pose calibration algorithm based on EKF by using the system information of current status to predict the next status system information.3. Design the measurement model of camera-IMU relative pose calibration algorithm based on EKF by using the coordinates of the image feature points in2D image coordinates system as the EKF measurement vector, computing the partial derivative of the system status vector and the Jacobian matrix of the measurement model formula. Using this information to correct the system status vector estimated value iteratively.4. Construction the camera-IMU relative pose calibration platform, generate simulated data, using simulated data to validate the availability and correctness of the algorithm introduced.