Development of a hashing-based data structure for the fast retrieval of 3D terrestrial laser scanned data

The volume of point cloud data obtained by 3-dimensional terrestrial laser scanners has grown very large as a result of scanner enhancements and application extensions. Quick point querying is therefore essential for efficient point cloud processing, and several data structures are applicable for that purpose. Octree, for example, is utilized in similar approaches and is considered a good candidate. This paper introduces hashing-based virtual grid (HVG), both as a competitor for octree and an improvement on the 3-dimensional virtual grid (3DVG). Whereas 3DVG is defined as a 3-dimensional array, HVG substitutes hashes for 3DVG's vertical indices. The performance of HVG was evaluated against those of octree and 3DVG by a point-querying operation. The selected operation finds neighboring points residing within a given radius for every individual point in the point cloud. HVG proved its balancing aspects throughout the operation, showing reasonable performance and memory efficiency. 3DVG, while its performance was excellent, required a significantly larger amount of memory. In summary, HVG is a suitable alternative to octree, and is expected to be effectively utilized as a base data structure for any application dealing with a massive amount of 3-dimensional point cloud data.

► We introduce hashing-based virtual grid (HVG) as a large 3D point cloud indexing structure.
► The performance of HVG is evaluated against octree and 3D virtual grid (3DVG) by a query operation.
► HVG proves its balancing aspects of reasonable performance and memory efficiency.
► HVG is a suitable alternative to octree and 3DVG.