The visual inertial navigation system calculates the initial pose,inertial measurement unit(IMU)bias,and monocular scale factor during the initialization phase,which can greatly ensure system convergence.To address the issue of poor or even failed initialization of visual inertial navigation systems in environments with missing point features,an initialization method for visual inertial navigation system based on point line is proposed.Firstly,a point line based structure is constructed within the sliding window using geometric methods,which has richer prior information compared to point structures.Secondly,the method of restoring structure from motion(SfM)is used to restore the motion state of the point line structure.Finally,align the motion state of the structure and the IMU pre integration results in a loose combination form,and solve for the initial variables such as IMU bias,gravitational acceleration,velocity,and scale factor.The experimental results show that on the public dataset with missing point features,compared to the original algorithm,the proposed method reduces the root mean square localization error by 6%,significantly reduces the initial error,and has better mapping results.
万方数据
维普
