Evaluating the benefits of Octree-based indexing for LiDAR data

In recent years the geospatial domain has seen a significant increase in the availability of very large three dimensional (3D) point datasets. These datasets originate from a variety of sources, such as for example Light Detection and Ranging (LiDAR) or meteorological weather recordings. Increasingly, a desire within the geospatial community has been expressed to exploit these types of 3D point data in a meaningful engineering context that goes beyond mere visualization. However, current Spatial Information Systems (SISs) provide only limited support for vast 3D point datasets. Even those systems that advertise their support for in-built 3D data types provide very limited functionality to manipulate such data types. In particular, an effective means of indexing large 3D point datasets is yet missing, however it is crucial for effective analysis. Next to the large size of 3D point datasets they may also be information rich, for example they may contain color information or some other associated semantic. This paper presents an alternative spatial indexing technique, which is based on an octree data structure. We show that it outperforms R-tree index, while being able to group 3D points based on their attribute values at the same time. This paper presents an evaluation employing this octree spatial indexing technique and successfully highlights its advantages for sparse as well as uniformly distributed data on the basis of an extensive LiDAR dataset.