outline
play

Outline Introduction & Related Works Overview User Interface - PDF document

2013/10/24 IEEE Pacific Visualization Symposium 2011 TripVista: Triple Perspective Visual Trajectory Analytics and Its Application on Microscopic Traffic Data at a Road Intersection Hanqi Guo 1, 2 , Zuchao Wang 1 , Bowen Yu 1 , Huijing Zhao 1 ,


  1. 2013/10/24 IEEE Pacific Visualization Symposium 2011 TripVista: Triple Perspective Visual Trajectory Analytics and Its Application on Microscopic Traffic Data at a Road Intersection Hanqi Guo 1, 2 , Zuchao Wang 1 , Bowen Yu 1 , Huijing Zhao 1 , Xiaoru Yuan 1, 2 1) Key Laboratory of Machine Perception (Ministry of Education), and School of EECS 2) Center for Computational Science and Engineering Peking University, Beijing, P.R. China Outline • Introduction & Related Works • Overview • User Interface • Results • Conclusion 1

  2. 2013/10/24 Introduction – Traffic Management • Traffic management has always been a critical issue in modern society i i d i t A traffic jam on National Expressway 110 in China, captured at Aug. 24 th , 2010, which stretched for 60-mile, lasted 11 days http://www.foxnews.com/world/2010/08/24/long-haul-chinas-traffic- http://www.news-to-use.com/2010/08/china- jam-stretching-long-km-weeks/ traffic-jam-could-last-weeks.html Introduction – Research on Traffic • Scope – Macro-trend of traffic – Micro-behaviors of traffic • To find out origin of accidents and jams • To evaluate traffic light configurations • Data source – Simulation data – Real data • Many features and exception are not well-modeled M f t d ti t ll d l d • Suffer from noise • Challenges – Complexity, Noise, Size 2

  3. 2013/10/24 Related Works • Spatial generalization and aggregation [Andrienko2011] Related Works • Spatial, temporal, attributes information closely linked l l li k d FromDady: [Hurter2009] 3

  4. 2013/10/24 Overview – Data Acquisition • Laser scanning Raw data collected as point cloud Preprocessed data as moving object, further classified Image courtesy of Zhao et al. 2009 Overview – Data Acquisition • Raw data as point cloud Video courtesy of Zhao et al. 2009 4

  5. 2013/10/24 Overview – Data Description • Microscopic trajectory dataset collected at a road intersection intersection – 209,426 trajectories • Length, width • Track type: pedestrian, bus, bicycle, car and other • Sample point array – 33,362,651 sampled points 33 362 651 l d i t • Position, speed, direction, time – 2 days’ data: 7:00 ~ 21:00 – Noise Overview – Design Philosophy • Triple perspectives closely linked 5

  6. 2013/10/24 User Interface • Spatial view + Temporal view + PCP Spatial View Temporal View Control Panel Time range selection Parallel Coordinates Spatial View Compass • Spatial perspective – Trajectory as curve – Compass 6

  7. 2013/10/24 Spatial View • Spatial perspective – Trajectory as curve – Compass – Ring-slider Entrance Angle Filter Exit Angle Filter Spatial View Exit Angle Histogram • Spatial perspective – Trajectory as curve – Compass – Ring-slider – Histogram Entrance Angle Histogram 7

  8. 2013/10/24 Temporal View • Temporal perspective – ThemeRiver h • Glyph embedded – Scatterplots • Start time VS Passing time • Start time VS MinSpeed • Start time VS AvgSpeed • Start time VS MaxSpeed • Start time VS Distance Temporal View • Glyph design – Trajectory direction clustering and glyph design Ordinary directions Outlier 8

  9. 2013/10/24 Temporal View • Algorithm for ThemeRiver with glyphs – Original ThemeRiver [Harve2002] Temporal View • Algorithm for ThemeRiver with glyphs – Original ThemeRiver [Harve2002] – Possible glyph position • Fast Hierarchical Importance Sampling [Ostromoukhov2004] 9

  10. 2013/10/24 Temporal View • Algorithm for ThemeRiver with glyphs – Original ThemeRiver [Harve2002] – Possible glyph position • Fast Hierarchical Importance Sampling [Ostromoukhov2004] – River subdivision Temporal View • Algorithm for ThemeRiver with glyphs – Original ThemeRiver [Harve2002] – Possible glyph position • Fast Hierarchical Importance Sampling [Ostromoukhov2004] – River subdivision – Final result 10

  11. 2013/10/24 Parallel Coordinates • Multi-dimensional perspective – Two raw dimensions: start time, type d – Nine derived dimensions: total time, average speed, minimum speed, maximum speed, total distance, beginning angle, ending angle, angle change, smoothed angle change, minimum acceleration, maximum acceleration Positive Negative Correlation Correlation User Interactions 11

  12. 2013/10/24 Results • Case I: Investigate specific behaviors • Case II: Find patterns and violations • Case III: Discover hidden information Case I – Investigate Specific Behaviors • Recognize special spatial patterns in the traffic view Turn-around pattern d 12

  13. 2013/10/24 Case I - Investigate Specific Behaviors (cont’d) • By ring-sliders Case I - Investigate Specific Behaviors (cont’d) • By sketch 13

  14. 2013/10/24 Case II – Find Patterns and Violations (cont’d) • Micro behavior: violation detection Case II – Find Patterns and Violations (cont’d) • Pattern discovery – Temporal periodicity due to traffic light regulation T l i di it d t t ffi li ht l ti – Volume comparison, may guide the optimization of traffic light control 14

  15. 2013/10/24 Case III – Discover Hidden Information • Discover exceptional trajectory cases by multi- di dimensional analysis i l l i – Angle change of the trajectory may reflect some dangerous cases Angle Change Case III – Discover Hidden Information (cont’d) 15

  16. 2013/10/24 Case III – Discover Hidden Information (cont’d) Case III – Discover Hidden Information (cont’d) • Similar patterns detected by the relative motion detection algorithm ti d t ti l ith Jun 16 th , 11:22am Jun 16 th , 12:06pm Jun 16 th , 3:12pm Jun 22 nd , 7:27pm Jun 22 nd , 6:54pm However, when speed is considered, none of them are really dangerous. 16

  17. 2013/10/24 Results • Case I: Investigate specific behaviors • Case II: Find patterns and violations • Case III: Discover hidden information Conclusions • TripVista, a visual analytics system for microscopic traffic trajectory data i i t ffi t j t d t • A coordinated visualization based on the triple perspective design philosophy • Capacity to reveal traffic patterns as well as abnormal behavior abnormal behavior 17

  18. 2013/10/24 Future Works • Incorporate more automatic algorithms • Extend our system to more complex trajectory dataset, e.g. road network • Integrate videos to the system Acknowledgements • Funding: – NSFC No. 60903062, BNSF No. 4092021, 973 No.2009CB320903, 863 NSFC No. 60903062, BNSF No. 4092021, 973 No.2009CB320903, 863 2010AA012400, MOE Key Project No. 109001 and IIPL-09-016 • Anonymous reviewers for comments • Jie Sha for data preprocessing and feedbacks http:// vis.pku.edu.cn 18

Recommend


More recommend