Activity Analysis in Wide-Area Aerial Surveillance Video
Abstract
In this work, we present methods for tracking and activity analysis that are scalable and appropriate for challenges present in wide-area aerial video. After an initial preprocessing stabilization step, we use a constrained interest point matching algorithm to generate weak tracks of vehicles in the scene. We then present algorithms that exploit the tracks to recognize traffic activity patterns of flow direction, uni vs. bidirectional traffic, acceleration/deceleration zones, and bidirectional stops via analysis of characteristic velocity patterns. Lastly, we provide quantitative and qualitative results of our activity analysis algorithms using both synthetic and real imagery.
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 01, 2011
- Accession Number
- AD1001185
Entities
People
- James W. Davis
- Ryan Feather
Organizations
- Ohio State University