Your SlideShare is downloading. ×
Gis cience 2010 presentation
Upcoming SlideShare
Loading in...5
×

Thanks for flagging this SlideShare!

Oops! An error has occurred.

×
Saving this for later? Get the SlideShare app to save on your phone or tablet. Read anywhere, anytime – even offline.
Text the download link to your phone
Standard text messaging rates apply

Gis cience 2010 presentation

419
views

Published on

How to detect stops in GPS data? An exploratory spatial data analysis approach

How to detect stops in GPS data? An exploratory spatial data analysis approach

Published in: Technology

0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total Views
419
On Slideshare
0
From Embeds
0
Number of Embeds
1
Actions
Shares
0
Downloads
0
Comments
0
Likes
0
Embeds 0
No embeds

Report content
Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
No notes for slide
  • Good morning. My name is Daniel Orellana from Wageningen Universityand I will present the results of our research on uncovering patterns of movement suspension.
  • Those places usually are associated to geographical features that can be related to specific activities. Therefore, we can extract useful knowledge about the interactions between the moving entities and the environment by analysing the places where they stop. The general idea is, paraphrasing Helen Couclelis: Tell me where they stop and I will tell you what (most probably) they are doing.
  • However, finding those places may be cumbersome: There are importantlimitations depending on the approach used to detect the stops. Most of the approaches reported in the literature requires a parameterisation of spatial or temporal thresholds to define the stops. For example, a stop may be conceptualised as when an object remains inside a predefined region for a minimum time; when an object does not move more than a predefined distance for a minimum time, or when the speed of the object remains below a threshold also for a predefined time.Defining those thresholds depends on the application and the kind of moving objects. For example for vehicles moving in a street network it would be relatively easy to define a speed threshold below which the vehicle is considered as stopped.
  • But when we analyse the movement of other kind of objects, such as pedestrians or animals, Other limitations related to the observation techniques will arise. Most of the movement data comes from positioning datasets collected with GPS. Due to intrinsic inaccuracy of GPS signal, there are not observations of speed equal to zero. Moreover, may be difficult to distinguish slow movement from measurement errors.
  • Taking into account these limitations, we are investigating a method to detect patterns of movement suspension able to work for different kinds of moving objects (even slow as pedestrians); independent of the application (no need of spatial or temporal thresholds) and independent of the scale.Developing a method to detect suspension patterns for movement of collectives using the spatial association of movement vectors.
  • In order to test the approach, we used the LISA index to find suspension patterns in four data sets representing the movement of different kind of objects.In the following slides I will present the results for each one.
  • The first dataset represents the movement of four hundred nineteeen children playing a urban mobile game and consisted of sixty one thousand movement vectors collected each 10 seconds for 6 days. Applying the LISA approach, we detected 55 clusters of movement suspension. The larger clusters were located at the checkpoints of the game where the players must go to win points. Other clusters were associated with game events such as encounters of players during the game. Interestingly, we also found two suspension patterns associated to pedestrian crossings in busy streets.
  • The second dataset represents the movement of three hundred seventy two visitors in the Dwingelderveld National Park in The Netherlands. More than one hundred forty one thousands vectors collected with variable time rate for fourteen days. Here we detected one hundred fifty two clusters of movement suspension. The largest clusters are associated to the parking lots where the entrances of the park are located. Other clusters were associated to the main attractions and facilicities in the park and many small clusters were located in the cross paths, probably indicating a route choosing behaviour.
  • The next data set represents the movement of 50 trucks delivering cement in Athens. It consists on one hundred eleven thousands movement vectors collected every thirty seconds for forty days. Here we found 252 clusters. The two largest clusters contained more than thirty per cent of the vectors and were located at the distributions points. Other large clusters were located at important building projects such as the Olympic Village (Figure. 3a) and the AttikiOdos Avenue (Figure. 3b). Other clusters with elongated shape may represent slow movement due traffic jams.
  • The last data set, represent the movement of five elephants in the Kruger National Park in South Africa. The movement of the animals was tracked each hour for thirty months collecting a total of one hundred thousand movement vectors. In this data set we detected eighty five clusters, mainly located close to river banks, bodies and fenced places representing obstacles for the movement.
  • In this overview, you can see the proportion of vectors classified as suspension and the number of clusters for each experiment
  • We also evaluated the results in terms of how many suspension patterns are associated to relevant places. In the Dwingelderveld National Park, for example, over the ninety per cent of the suspension patterns are associated to relevant places. We consider the other nine per cent as false positives.
  • We are investigating three limitations related to our approach. The first is that is not able to detect suspension patterns for different entities moving together.The second one, is related with the size of the data set, because it is possible that the local spatial association may be obfuscated in huge data sets.The third limitation is that the approach is based on the statistical properties of the whole data set and may be not suitable for real time analysis.
  • Finally our future work will be focused on the contextualisation and interpretation of the suspension patterns using the characteristics of the moving entities
  • Transcript

    • 1. Uncovering patterns of movement suspension
      D. Orellana, M. Wachowicz, H.J. De Knegt, A. Ligtenberg and A. K. Bregt
      Centre for Geo-Information
      Wageningen University
    • 2. Motivation
      One of the main tasks on movement analysis is to detect the places where moving entities stop.
      Suspension patterns represent the reduction of speed observed when moving entities stop.
    • 3. Limitations related with the approach
      Predefined regions
      Spatial thresholds
      Speed thresholds
      + Temporal thresholds
    • 4. Limitations related with observation
      • There are no observations of speed=0
      • 5. Slow movement may be undistinguishable from inaccurate observations.
    • Our approach
      • For different kind of moving objects.
      • 6. No need of spatial and temporal thresholds.
      • 7. Scale-independent.
      Developing a method to detect movement suspension patterns for collectives of objects using the spatial association of movement vectors.
    • 8. Different representations of movement
      Movement astrajectories
      Movement as vectors
    • 9. Spatial association of movement vectors
      A Local Indicator of Spatial Association (LISA) is used to find spatial clusters of low speed vectors.
      (LISA) cluster map for under-five mortality rate in Nigeria:O. A. Uthman 2008.
    • 10. Methodology
    • 11. Implementation
      Four datasets representing the movement of different kind of entities:
      Children playing an urban mobile game.
      Visitors walking in a recreational area.
      Trucks delivering cement in a city.
      Elephants moving in a Natural Park
    • 12. m
      - 419 children in Amsterdam
      - 61782 vectors
      - 10 secondsfor 6 days
      55 Clusters:
      • Checkpoints
      • 13. GameEvents
      • 14. Pedestriancrossings
      Urban Mobile Game
    • 15. - 372 pedestriansvisitingthepark
      - 141826 vectors
      - Variable time ratefor14 days
      Dwingelderveld National Park
      m

      152 Clusters:
    • Trucks in Athens
      [Theodoridis 2003]
      - 50 Trucks delivering cement
      - 111 419 vectors
      - 30 seconds for 40 days
      m

      z
      252 Clusters:
      • Distributionpoints
      • 19. Buildingprojects
      • 20. Trafficjams?
    • Elephants in South Africa
      [De Knegt, 2003]
      m

      z
      Å
      - 5 Elephants in South Africa
      - 100 337 vectors
      - 1 hourfor30 months
      85 Clusters:
    • Overviewof all experiments
      Proportion of vectors classified as suspension and number of spatial clusters of movement suspension detected in each experiment.
      Elephants
      7%
      85
      Trucks
      32%
      252
      Visitors
      6%
      152
      Children
      18%
      51
    • 23. Evaluation of the results
      In the Dwingelderveld National Park, over 90% of movement vectors were associated to relevant spatial features
      False Positives: 9.1%
      Unknown
      Cross paths
      Facilities
      Attractions
      Parking lots
      Movement
      Vectors
      Vectors classified
      as suspension
      100%
      80%
      True Positives: 90.9%
      60%
      40%
      20%
      0%
    • 24. m  z Å 
      Given a movement dataset, the suspension patterns are spatial clusters of movement vectors that fulfil three conditions simultaneously: (a) having speed values below the average for a given data set; (b) having a positive local spatial association of these speed values; and (c) having a minimum statistical significance score of this association corresponding to an established confidence level.
      • Different kind of objects moving in different landscapes
      • 25. Different spatial a temporal scales
      • 26. Without thresholds
    • Limitations
      • Different of moving entities together.
      • 27. Large datasets.
      • 28. Real time.
    • Future work
      Contextualisation and interpretation.
      Temporal association of suspension patterns.
    • 29. Thank you!
      Daniel Orellana
      Centre for Geo-Information
      Wageningen UR
      daniel.orellana@wur.nl
      http://wu.academia.edu/DanielOrellana

    ×