Digital terrain model generation using LiDAR Ground Points

As the trend of autonomous self-driving cars is becoming more of a reality, High-quality navigation methods and tools become a paramount. This, in turn, is crucially dependent on High-definition maps, for which one of the enabling tools is high resolution Digital Terrain Model (DTM) - the role and values of which have already been demonstrated even in the settings of manned cars. Traditional DTM generation methods have insurmountable barriers in creating centimeter-level resolution. In this paper, we propose a novel method for fully-automated, high precision DTM generation using the database generated and maintained in our existed dataset, and with no additional overheads in terms of extract labor and equipment cost. The input data is a point cloud captured by the vehiclemount LiDAR devices which, naturally, has extremely large volume. We show how with Ground Points Processing and DTM Generation steps, we can generate a centimeter-resolution DTM and, as our experiments demonstrate, when compared to DTM form U.S. Geological Survey (USGS) and altitude data from a third party surveying dataset, our proposed DTM indeed provides a higher precision.