In order to enable controllers to feel as if they are onsite when remotely controlling robots, the remote locations should be reproduced as three–dimensional (3D) landscapes. To facilitate this, the point cloud obtained from Light Detection And Ranging (LiDAR) sensor needs to be made into a 3D model. This point cloud comprises a number of points, which are classified into ground and non-ground datasets in order to effectively make 3D models that satisfy the features of each dataset. This paper proposes an approach that segments the non-ground dataset from a single scan of the point cloud of an unpaved road acquired via LiDAR.