| CPC G06V 20/58 (2022.01) | 6 Claims |

|
1. A trajectory correction method comprising the steps of:
correcting, by a map generation device, an error generated due to a change in position targeting 3D point cloud data collected in real time through a LiDAR, of which a position changes in real time, wherein the step of correcting the error includes:
correcting an error of the 3D point cloud data generated due to a change in the position based on inertial navigation system (INS) data and global positioning system (GPS) data collected together with the 3D point cloud data simultaneously;
generating fusion data complementing a relative position included in the INS data with an absolute position of the GPS data; and
searching for fusion data closest to a position of the LiDAR, and correcting the error of the 3D point cloud data generated due to the change in the position based on information on a time of collecting the 3D point cloud data and the fusion data;
estimating, by the map generation device, an expected trajectory (odometry) related to the change in the position based on the corrected 3D point cloud data, wherein the step of estimating the expected trajectory includes expressing the 3D point cloud data and the fusion data in a voxel space, and estimating the expected trajectory by approximating a point group included in each voxel using a Gaussian distribution;
applying, by the map generation device, curve fitting to the estimated expected trajectory; and
mapping, by the map generation device, the expected trajectory before the curve fitting targeting the expected trajectory to which the curve fitting is applied.
|