**3. The Preprocessing Module**

In this study, the current LIDAR point cloud is projected onto the current 2D image grid, which is represented by a matrix. The horizontal index unit of moment frame is the horizontal resolution of each frame, and the vertical unit is the vertical resolution. For example, the size of the projected image matrix of 16-line LIDAR is 16 ×1800. The value of the image grid stores the depth of each point, and the points will be removed if there is an outlier value. The operating point cloud data on the basis of two-dimensional images can significantly improve the computing speed.

After this process, reliable estimation of LIDAR´s per scan is a necessary prerequisite. In this paper, IMU pre-integration is used to obtain the relative translational motion at the beginning and the end of each scan. Based on this method, point cloud distortion can be eliminated. In the meantime, the raw point clouds from each scan are rotationally de-skewed using gyroscope data.
