The document proposes an incremental clustering algorithm to track objects consistently throughout a wireless sensor network. Static clustering is commonly used but has issues tracking objects at cluster boundaries. The proposed approach forms incremental clusters when objects are at boundaries to continue tracking in an energy efficient manner. It was found to be more stable and energy efficient than dynamic clustering approaches. The algorithm updates clusters incrementally in response to new object trajectory information.

1. Incremental Clustering-based
Object Tracking in
Wireless Sensor Networks
Presented by -
M S SACHIN

2. INTRODUCTION
• Amongst most of moving object tracking methods clustering
based object tracking has shown significant results,which in
term provides the network to be scalable and energy
efficient for large-scale WSNs
• static cluster based object tracking is the most common
approach for large-scale Wireless Sensor Network (WSN).
However, as static clusters are restricted to share
information globally, tracking can be lost at the boundary
region of static clusters.
• The main objective of this paper is an Incremental Clustering
Algorithm is proposed in conjunction with Static Clustering
Technique to track an object consistently throughout the
network solving boundary problem.

3. • Wireless Sensor Networks (WSNs) and their applications have
remarkable prospective in both military and commercial
places.
• Tracking objects throughout the established network is one of
the most concerned areas in WSNs. Battlefield surveillance,
emergency rescue, patient monitoring

4. Object tracking in WSNs involves
two steps
Step 1: Sensor nodes need to detect the presence of a moving
object within an area of interest, estimate its location, and
forward the location information to the base station.
Step 2: it needs to control the tracking process to monitor or
capture the moving object.

5. The WSN Tracking and Localization
Problem
• static clusters are bound to share information within cluster, so
tracking inconsistency may happen at the boundary during object
movement in boundary regions.
• energy cost gets higher for frequent cluster creation and deletion
during dynamic clustering.
• distance-based calculation in dynamic clustering has reduced the
reliability and accuracy in object localization and tracking.

6. Proposed technique for moving object
tracking techniques in WSNs.

7. Incremental Clustering-based Object
Tracking in Wireless Sensor Networks
• The main contribution of this paper is to propose an incremental
clustering algorithm for consistent object tracking at the boundary
region of static clusters
Principal Contributions :
• A complete and consistent algorithm for robust object tracking and
accurate localization in WSNs
• An energy efficient Incremental Clustering Algorithm for object
tracking at the boundary region of static clusters.

10. • Fig 2a: sensor nodes are partitioned into several static
clusters to continue energy-efficient tracking task with a low-
cost
• Fig 2b: clusters are represented as circular shape, solid lines
indicate active clusters and dashed lines indicate dismissal of
clusters.

11. • Fig 2c : When an object exits from the static cluster another
temporary cluster will be created and the process will continue
dismissing the previously created one.
• Fig 2d: Incremental clustering algorithm can create clusters
incrementally at the boundary region This technique can not
only create clusters, but also can retain recently created clusters
which in turn increase energy-efficiency.

12. A . Object Tracking and Localization in
WSNs Algorithm
• 1: Initialize Network with Static Cluster
• 2: Boundary Node Formation
• 3: while (object detect) do
• 4: if tracking is in safety region then
• 5: Tracking Continue with Static Cluster Representative
• 6: else (object is in boundary region) if Observation Trajectory
Satisfies Membership Condition of Incremental Clusters then

13. • 7: Update Clusters (Incrementally created) with new Observation
Trajectory
• 8: else Create a new Cluster using Incremental Clustering Algorithm
• 9: Tracking Continue with Increment Cluster Representative
• 10: end if
• 11: RSSI Analysis to Calculate Anchor Node from Selected Cluster
• 12: Localization using Trilateration Algorithm
• 13: end while

14. B. Incremental Clustering Algorithm -
1. cluster can grows incrementally. Network can retain
previously learned information which can solve
stability-plasticity dilemma.
• object tracking can continue smoothly in the border
region with energy-efficiency

15. 2. Clusters in a network follow Gaussian distribution As
sensors are deployed in 2D space Gaussian
distribution can be used.
A. Experimental Setup:
• Using Matlab Simulator, 100 nodes are deployed
randomly into the area of 100 x 100 m2 for moving
object monitoring
• Considering an application of radio transmission of
energy dissipation during transmit and receive,
assuming

16. (a) Deployment of Sensor Network and Static Clustering based Object
Detection.

17. B. Experimental Results -
• tracking sequences considering both Static Clustering
(LEACH protocol) and Incremental Clustering based
object tracking.
• indicate incremental clusters are formed when object
is in the boundary region of static clusters to
continue the task of tracking.

18. (b) Object Tracking Inside static Cluster using LEACH protocol.

19. (c) Incremental Clustering based
Object Tracking at the Boundary
Region of Static Clusters.
(d) Object Tracking Inside static
Cluster using LEACH protocol

20. Incremental Clustering vs. Dynamic
Clustering
0
20
40
60
80
100
120
0.0.0.0 2.5.0 5.0.0 7.5.0 1.0.0.0
Series 2
Series 1
Incremental
Clustering
Dynamic
Clustering
AliveNodes
Rounds

21. • that network is more stable and energy-efficient for
Incremental Clustering based tracking
• for choosing Dynamic Clustering average of 40 nodes
goes down
• only 12 nodes dies for choosing Incremental
Clustering at the boundary region of static clusters

22. CONCLUSIONS
• we have developed an approach for real-time
incremental learning of continuous flow of object’s
tracking
• The clusters are updated automatically
• In response to new information, the incremental
learning algorithm adapts to add this information