Automated Traffic Surveillance from an Aerial Camera Array

The overall goal of this O&E Grant is to design and develop an automated aerial network monitoring system concept that can identify and track individual vehicles through a network of 16 square miles in near real-time. The research includes the development of algorithms to map the locations of the vehicles and to extract traffic parameters for data mining purposes. Three (3) experimental vehicle tracking systems have been conceptualized and are being evaluated. They are: (1) Vehicle Tracking using a raw pixel appearance model/OpenStreetMap; (2) Surf feature tracking; and (3) Deep learning using the Caffe library (UC Berkeley). Based on results to date, the research team has selected a machine learning approach to detect and track vehicles in an aerial camera array video. The third system identified above, deep learning, appears more promising than other approaches--especially with challenging video sequences with seams and variation in luminance.


  • English


  • Status: Active
  • Contract Numbers:


  • Sponsor Organizations:

    Office of the Assistant Secretary for Research and Technology

    University Transportation Centers Program
    Department of Transportation
    Washington, DC  United States  20590
  • Performing Organizations:

    Clemson University

    College of Engineering and Science
    109 Riggs Hall, Box 340901
    Clemson, SC  United States  29631-0901
  • Principal Investigators:

    Sarasua, W

  • Start Date: 20140601
  • Expected Completion Date: 0
  • Actual Completion Date: 20150831
  • Source Data: RiP Project 40234

Subject/Index Terms

Filing Info

  • Accession Number: 01572494
  • Record Type: Research project
  • Source Agency: Southeastern Transportation Center
  • Contract Numbers: DTRT13-G-UTC34
  • Files: UTC, RiP
  • Created Date: Aug 8 2015 1:01AM