LiDAR data reduction for efficient high quality DEM production
Digital Elevation Models (DEMs) play an important role in terrain related applications. Generating DEM by traditional methods such as field surveying and photogrammetry is really time consuming and labour intensive. In densely forested area, it is impossible to use these methods for collecting elevation data. Light Detection and Ranging (LiDAR), an emerging technology, provides an alternative way to outcome these limitations. It offers capability of capturing three dimensional points with high density, enabling creation of detailed DEMs with high accuracy. The big challenge here is the processing of large volume of LiDAR data sets, resulting in extensive computational requirements for producing DEM. As a matter of fact, LiDAR may over-sample the terrain, particularly in flat areas, leading to data redundancy. To improve efficiency in both data storage and processing, the redundancy of terrain data must be minimized by eliminating unnecessary elements.
This paper presents ways to mitigate data redundancy in LiDAR-derived DEM by reducing LiDAR point density. Based on LiDAR data covering a total area of 6900 km2, in the Corangamite catchment region, south western Victoria, Australia, structure lines or breaklines such as ridge and valley lines which are crucial for the description of terrain surfaces are extracted from original dense LiDAR points (the highest available LiDAR point density for this study). These structure lines are then incorporated into sparse LiDAR point sets which are thinned out from the complete LiDAR data set to produce a new DEM with less data redundancy. In this way, the efficiency for data storage and processing is improved; meanwhile, the DEM generated from sparser data still preserve high accuracy.