Photo of Jan Prokaj, Ph.D.

Jan Prokaj, Ph.D.


Using 3D Scene Structure to Improve Tracking


Occlusion of vehicles in aerial imagery.
  • Aerial imagery collected over urban areas contains many occlusions for vehicles
  • Current tracking algorithms cannot track vehicles through such occlusions due to weak object appearance and complexity of motion prediction over long period


Flowchart describing the proposed approach.
  • Use known 3D scene structure to estimate a dynamic occlusion map: binary map, which indicates what regions of the image are occluders of moving objects; dynamic due to camera motion
  • Occlusion map.
  • Cluster track initiations and terminations near an occlusion map to find sources and sinks
  • Find corresponding source-sink pairs using the Hungarian algorithm
  • Match a sequence of tracks at the sink with a sequence of tracks at the source using dynamic programming
  • Merge the matched tracks


Proposed AlgorithmHungarian Algorithm
Correctly Linked58%42%
Incorrectly Linked3.4%5.8%




J. Prokaj , G. Medioni . Using 3D Scene Structure to Improve Tracking. In IEEE Conference on Computer Vision and Pattern Recognition , pages 1337-1344, 2011.


Related Publications

  • J. Prokaj . Exploitation of Wide Area Motion Imagery . Ph.D. thesis, University of Southern California, 2013.