Augmented Reality Based On A Mixture Of Visual And Inertial Sensors To Follow-up Study

Augmented Reality (AR) has been produced a new Computer Application and Human-Computer Interaction technology on the basis of the Virtual Reality, adding virtual computer-generated material to the surrounding physical world by tracking the user's pose in the real scene, in order to improve the user's perception. Registration technology, virtual-reality registration by gaining the pose information of the users with respect to the real scene, plays an important role in the AR system.For visual-based tracking system, the paper discusses and carries out the selection and identification of the Marker, camera calibration and the virtual-reality registration. The experimental analysis shows: the visual-based tracking system fails to track when the camera moves too fast or the Marker is obscured. For the above problems, this paper presents a vision-based and inertial sensor data fusion method to obtain the pose information of the users with respect to the real scene when the camera undoes it. After the calibration between the visual and inertial sensor, establishing the time dynamic model of the orientation and position. Extended Kalman filter (EKF) is used to amend and estimate the orientation of the visual sensor with respect to the Marker, and discrete Kalman filter (DKF) is applied to amend and estimate its position. By data and image comparison, it not only improves robustness and rapidity of the visual-based tracking, but also the accuracy of the inertial-based tracking.Finally, this paper introduces the visual-based tracking and mixed-tracking respectively in the equipment maintenance AR system. Through both visual effects and experimental data, the latter solves the defect of the former, obtaining the better experimental effectiveness.