Authors:
Léonie Buckley
;
Jonathan Byrne
and
David Moloney
Affiliation:
Intel, Ireland
Keyword(s):
Hashing, SLAM, Storage, Indexing, Volumetric Data, Primes, Voxels.
Abstract:
Traditional hashing methods used to store 3D
volumetric data utilise large prime numbers in an attempt to
achieve well-distributed hash addresses to minimise addressing
collisions. Hashing is an attractive method to store 3D volu-
metric data, as it provides simple method to store, index and
retrieve data. However, implementations provide no evidence
as to why they utilise large primes [1], [2], [3], [4], [5], which
act to create a hash address through randomising key values.
The coordinates of quantised 3D data are unique. It is thus
investigated in this paper whether whether this randomisation
through the use of large primes are necessary. The history of
the use of primes for hashing 3D data is also investigated, as
is whether their use has persisted due to habit rather than due
to methodical investigation.