1. Shape Segmentation and Applications in Sensor Networks
Xianjin Zhu Rik Sarkar Jie Gao
INFOCOM 2007
Lee, sunkyung

2. Contents Introduction Background Algorithm Outline
Segmentation algorithm
Application
Conclusion
ⓒ KAIST SE LAB 2008

3. Introduction(1/6) Common assumption
Sensor nodes are deployed uniformly inside a simple geometric region
e.g) square
In practice, can be complex
Obstacles (lakes, buildings)
Terrain variation
irregular sensor field !
ⓒ KAIST SE LAB 2008

4. Introduction(2/6) Some protocols may fail
Greedy forwarding : packets are
greedily forward to the neighbor
closest to the destination
Dense uniform
Sparse, non-uniform
Works well
May get stuck
ⓒ KAIST SE LAB 2008

5. Introduction(3/6) Some protocols have degraded performance
Quad-tree type data storage hierarchy
Data is hashed uniformly to the quads
ⓒ KAIST SE LAB 2008

6. Introduction(4/6)
Quad-Tree Type Hierarchy
ⓒ KAIST SE LAB 2008

7. Introduction(5/6) Motivation
Global geometric features affect many aspects of sensor networks
Affect system performance
Affect network design
Place base stations and
avoid traffic bottleneck
ⓒ KAIST SE LAB 2008

8. Introduction(6/6) Research goal : Shape segmentation
Segment the irregular field into “nice” pieces
Each piece has no holes, and has a relatively nice shape
Apply existing protocols inside each piece
Existing protocols are reusable
ⓒ KAIST SE LAB 2008

9. Background(1/4) Segmentation with Flow Complex
Flow complex in continuous domain
Distance function : h(x) = min{║x - p║2 : p on boundary}
Medial axis : a set of points with at least two closest points on
the boundary
Local max
H(s1) s1 s2 s3 p1
H(p1)
Reference : flow complex [Dey, Giesen, Goswami, WADS’03]
ⓒ KAIST SE LAB 2008

10. Background(2/4) Segmentation with Flow Complex(cont’d)
Flow complex in continuous domain
Flow direction: the direction that h(x) increases fastest
Sinks: local maximum, no flow direction
H(p1) p1 p2
Local max s2 s1 s3
Reference : flow complex [Dey, Giesen, Goswami, WADS’03]
ⓒ KAIST SE LAB 2008

11. Background(3/4) Segmentation with Flow Complex(cont’d)
Flow complex in continuous domain
Flow direction: the direction that h(x) increases fastest
Sinks: local maximum, no flow direction
Segments : set of points flow to the same sink
Local max s1 s3
Naturally
partition
along
narrow
necks s2
Reference : flow complex [Dey, Giesen, Goswami, WADS’03]
ⓒ KAIST SE LAB 2008

12. Background(4/4) Implementation Challenges
No global view, no centralized authority
No location, only connectivity information
Distances are approximated by hop count
Goal : A distributed and robust segmentation algorithm
ⓒ KAIST SE LAB 2008

13. Algorithm Outline
ⓒ KAIST SE LAB 2008

14. Segmentation Algorithm(1/7)
Step 1 : Compute the medial axis
Boundary nodes flood inward simultaneously
Nodes record : minimum hop count &
closest intervals on the boundary
Medial axis : more than two closest intervals
ⓒ KAIST SE LAB 2008

15. Segmentation Algorithm(2/7)
Step 2 : Compute the flow
Flow direction : a pointer to a neighbor with a higher
hop count from the same boundary
Prefer neighbor with the most symmetric interval
Sinks must be on the medial axis
Nodes are classified into segments by their sinks
Too many segments!
ⓒ KAIST SE LAB 2008

16. Segmentation Algorithm(3/7)
Step3 : Merge nearby sinks
Nearby sinks with similar hop count to the boundaries are merged (together with their segments)
ⓒ KAIST SE LAB 2008

17. Segmentation Algorithm(4/7)
Step3 : Merge nearby sinks(cont’d)
Nearby sinks with similar hop count to the boundaries are merged (together with their segments)
Segmentation granularity : |Hmax – Hmin| < t
ⓒ KAIST SE LAB 2008

18. Segmentation Algorithm(5/7)
Step4 : Final clean-up
Orphan nodes
Local maximum and nodes that flow into them
Merge orphan nodes with nearby segments
ⓒ KAIST SE LAB 2008

19. Segmentation Algorithm(6/7)
Final result
ⓒ KAIST SE LAB 2008

20. Segmentation Algorithm(7/7)
Simulation Results on Segmentation
ⓒ KAIST SE LAB 2008

21. Application(1/2) Distributed index for multi-dimensional data(DIM)
Suffer from a complex shape sensor field
Shape segmentation can reduce load and communication cost
before segmentation
after segmentation
storage load
Cost per event insertion
cross
corridor
Fish
Without shape segmentation
293.69
359.21
254.37
With shape segmentation
84.15
151.05
204.58
ⓒ KAIST SE LAB 2008

22. Application(2/2) Random sampling
Shape segmentation can reduce the number of trials and cost
No. of trials
cross
corridor
Fish
Without shape segmentation
168
149
136
With shape segmentation
112
115
123
Cost per sampling
cross
corridor
Fish
Without shape segmentation
477.84
511.95
361.80
With shape segmentation
102.49
182.32
238.47
ⓒ KAIST SE LAB 2008

23. Conclusion Contribution Future work
A unified approach handling complex shape in sensor networks
A good example to extract high-level geometry from connectivity information
Network self-organizes by local operations
Future work
Find the definition to choose appropriate segmentation
It depends on application
ⓒ KAIST SE LAB 2008

24. Question ?
ⓒ KAIST SE LAB 2008