Processing ObjectOrientationbased Direction Queries in SDB - PowerPoint PPT Presentation
1 / 13
Title:
Processing ObjectOrientationbased Direction Queries in SDB
Description:
8/31/09. 8705 Scientific Databases G8. 1 ... Compose, reverse: produce new directions. Deviation: calculate deviation between two directions ... – PowerPoint PPT presentation
Number of Views:21
Avg rating:3.0/5.0
Title: Processing ObjectOrientationbased Direction Queries in SDB
1
Processing Object-Orientation-based Direction
Queries in SDB
- Alex Zhang
- changQing Zhou
- Group 8
2
Outline
- Problem introduction
- Problem formulation
- Major contribution
- Key concept
- Validation
- Assumptions
- Future work
3
Problem Introduction RQS_AD
- Figure 4
- Query List all objects that are exactly east of
object B - Exercise
- Use direction rectangle
- Limitation What if the orientation of the
reference objects is not aligned with the global
coordinate system
4
Problem Introduction RQS
- Figure 5
- Query List all objects that are exactly in front
of object B - Exercise
- Use direction MBR
- Limitation direction MBR is bigger than
direction rectangle, large intermediate result
set, post filter required
5
Problem Statement
- Given
- A query involving object-orientation-based
directions - A R-tree index of the dataset
- Find All objects satisfying the selection
condition - Exercise
- Objective Minimize the number of page access
- Constraint
- Objects are approximated as MBRg
- There is only one R-tree index based on global
coordinate system
6
Major Contribution
- New ADT for open shapes
- New strategy Open shape-based strategy(OSS)
- Do not need to know the world boundary
- Reduce unnecessary I/O and CPU cost
- Eliminate PostFilter step for MBR objects
- Algebraic analysis of relative performance
- Experimental evaluation
- OSS outperforms RQS in both I/O and CPU cost
7
Key Concept Modeling Directions
- Model direction as an object vector
- Basic operations
- Compose, reverse produce new directions
- Deviation calculate deviation between two
directions - isBetween relationships among directions
- Orientation O ltOP, right, frontgt where OP is
the original point and right, front are two
orthogonal directions
8
Key Concept OSS
- Figure 5
- Use actual direction region as the filter
- Improve filter efficiency by eliminating false
hits at the earliest opportunity while
recursively searching R-tree - OSS is correct and complete
9
Key Concept OpenShape
- Figure 8
- Geometries whose boundary are partially defined.
One side open and two side open - Process direction predicates using topological
relationship between MBR objects and OpenShapes
10
Key Concept interior intersects
- Figure 9
- Figure 10
- O.interiorIntects( C ) TRUE ?
- Interior( C ) interior( O ) ! 0
- InteriorIntercts is correct and complete
11
Validation
- Strategy OSS and RQS
- Dataset
- Randomly generated data sets
- Randomly generated reference object sets
- Variable parameters
- Orientations of reference objects
- Size of reference objects
- Overlap degree of the dataset
- Number of objects in the dataset
- Results
- Number of disk pages accessed
- Number of operations performed
12
Assumptions
- Objects are approximated by MBR
- One R-tree index based on global coordinate
system no multiple R-tree indices with
different orientation
13
Future Work
- Modeling directional realtionships between
mix-dimension objects - Query processing and optimization
- Other spatial index methods
- Join on direction predicates
Recommended
CrystalGraphics Presentations
