incident identification platform

My presentation for the PATW competition 2013. A prototype platform that identifies incidents by accumulating reports from mobile devices.

Transcript

1. Incident Identification Platform Based
   on Mobile devices
   Anastasios Latsas
   IET - PATW 2013
   

   View Slide

2. Importance of Incident
   Detection
   

   View Slide

3. Incident Detection Systems
   TraCS
   (Traffic and Criminal Software)
   

   View Slide

4. Incident Detection Systems
   SARACEN
   (Socially Aware, collaboRative, scAlable Coding mEdia distributioN)
   

   View Slide

5. Incident Detection Systems
   Delasoft Incident Locator Tool
   

   View Slide

6. Motivation
   ● Existing systems are based on user input
   (assume location knowledge)
   ● Not "smart" enough
   ● Not open
   ● Not accessible
   

   View Slide

7. Incident Identification Platform
   ● Detect incidents without assuming any location
   knowledge
   ● Detect both static and evolving incidents
   ● Web based platform (accessible)
   ● Mobile clients for Android™ devices
   ● Communication using open APIs and common
   protocols (HTTP)
   ● Open source
   

   View Slide

8. Static Incidents
   1. calculate all possible incidents using
   the collected reports
   2. filter out too distant incidents
   3. calculate the average incident
   location
   4. calculate standard deviation of
   incidents and the average location
   5. determine safe point of location
   identification
   

   View Slide

9. Evaluation
   ● collecting real data (using different
   mobile devices)
   ● collecting virtual data (using gmaps)
   ● randomizing collected data
   (location)
   ● randomizing collected data (run
   order)
   ● replaying scenarios
   ● plotting results
   

   View Slide

10. Static Incidents - results
    

    View Slide

11. Detection Algorithm
    Tracking
    Algorithm
    ● 3 main phases
    ● filtering based on time and
    space criteria
    ● can detect evolving incidents
    ● adjusting existing position using
    two different algorithms
    ● scalable
    

    View Slide

12. Static vs Tracking
    algorithm performance
    

    View Slide

13. Future work
    Expand available clients
    

    View Slide

14. Thank You
    @tlatsas
    tlatsas@gmx.com
    Acknowledgements:
    Many thanks to Prof Ch. Patrikakis for his support in the
    design and implementation of the idea
    

    View Slide