| With the popularity of smart mobile device, the application of Augmented Reality Technology(AR) have got more rising space. On the condition that current 3D registration technologies for AR mostly rely on artificial markers or CAD models which makes their robust limited by graphics quality. This article proposed an autonomic SINS/SLAM combination positioning method aimed at continuously estimating metric position and pose in an unknown scene. In the monocular SLAM algorithm, the image characteristics of environment are extract for producing a 3D map of point features and tracking erratic hand-held motion robustly. The main research work includes:1. Proposed a SLAM/SINS combination positioning method aimed at continuously estimating metric position and pose in an unknown scene. Monocular SLAM is regarded as a black box and outputs camera pose, which reduce the computational cost of EKF SLAM from O(M3) to O(M2) at its theoretical best, with m being the amount of features.2. Research on monocular SLAM based on natural characteristics. This article divides traditional SLAM into two parallel threads.one thread deals with the task of robustly tracking erratic hand-held motion, while the other produces a 3D map of point features from previously observed video frames. The map is densely initialized from a stereo pair (5-Point Algorithm) and Mapping is based on keyframes, which are processed using batch techniques (Bundle Adjustment).3. Achieved a extend Kalman filter(EKF) which combines data from SLAM and SINS. The output of SLAM is regarded as the observed quantity in EKF, while the measurement of SINS is regarded as the quantity of state. When the device is static, the EKF can be used to demarcate the gravity component in the output of accelerometer.4. Conducted experiments which prove the positioning scheme is reliable. The filter convergence experiment, static stability experiment and displacement estimation experiment for SLAM/SINS combination system are executed separately. |
PDF Full Text Request