Spatial Indexing of Large Volume Bathymetric Data

1
Spatial Indexing of Large Volume Bathymetric
Data
by Bradford G. Nickerson University of New
Brunswick Faculty of Computer Science Fredericton,
New Brunswick Canada joint work with Feng Gao
2
Bathymetric Data Observation
Simrad ME70 transducer array
Frequency range 70 to 120 kHz 800 Tx 800
Rx channels
3
Vessel Track
4
Profiles
5
Edit Gridding?Contours
6
Navigation Errors Oops!
7
(No Transcript)
8
Experimental DataConceptionBay,
Nfld.57,192profilesin 49 lines130 MB
9
Summary

• R-tree average 4.6 times less time to search
  R-tree indices (compared to Morton code)
• Space requirement about 1.8 of original
  data size
• Range deletion average 17 times faster
  compared to original R-tree deletion
• RDBMS (INGRES) took an average of 1,535 times
  more CPU time for search, 4.5 times more
  space

10
Curse of Dimensionality

• k-d tree range search time O(n1-1/dF), where F
  is the number of points in range Lee and Wong,
  1977

For uniform, random points in 0,1d, query
square W with side length ?, ? d-th root of
E(F)/n. e.g. E(F)/n0.001, for d2,
?0.0011/20.03 for d20, ?0.0011/200.79
11
Sensor Web Language (SWL) Software Architecture
12
(No Transcript)