HOME MyTRB CONTACT US DIRECTORY E-NEWSLETTER FOLLOW US RSS


The National Academies

NCHRP IDEA 20-30/IDEA 140 [Completed (IDEA)]

Computer Vision Traffic Sensor for Fixed and Pan-Tilt-Zoom Cameras
[ NCHRP 20-30 (NCHRP-IDEA) ]

  Project Data
Staff Responsibility: Dr. Inam Jawed

This project involved development and testing of a next generation vision-based traffic sensor to collect traffic parameters such as volume, classification, and speed using both fixed and pan-tilt-zoom (PTZ) cameras. The developed sensor is quick and easy to calibrate using just six clicks in the image and also has the ability to dynamically recalibrate itself when the camera undergoes PTZ changes. Two prototype sensors were deployed at two locations (Maryland and New York) and have been under testing for more than 15 months. The sensor in Maryland has been operational continuously since operation while that in New York rather intermittently due to frequent loss of solar power. Due to the amount of manual labor involved in generating ground-truth data for classification and speed, evaluation was mainly focused on comparing vehicle counts. The sensor’s accuracy in terms of vehicle count was found to be comparable under various traffic, weather, and lighting conditions and often slightly better that that of the loop detectors present at the corresponding sites.  The improvement was particularly noticeable in congested traffic conditions encountered at the New York test site. The project has helped make significant progress toward developing a Traffic Management Center (TMC) solution using existing pan-tilt-zoom cameras. The architecture of the software was redesigned to enable the processing of multiple (up to 32) videostreams simultaneously on a single server. Video data from local TMC in Greenville, SC has been collected for development and testing purposes. An automatic calibration algorithm to handle user pan and tilt has been developed and demonstrated on several occasions. Once issues regarding network security have been addressed, the sensor will be installed at the TMC in Columbia, SC, thus opening the door for further development. A patent for the developed sensor technology has been filed. A local software company is collaborating in taking the IDEA product to the market. The contractor's final report is available.

To create a link to this page, use this URL: http://apps.trb.org/cmsfeed/TRBNetProjectDisplay.asp?ProjectID=2700