Accuracy of a stereo vision system for portable roadside portal monitoring

R. A. Kerekes, J. S. Goddard, A. M. Cheriyadat, T. P. Karnowski, M. F. Cunningham, D. E. Hornback, L. Fabris, K. P. Ziock

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Increasing concern over the smuggling of fissile material has prompted recent interest in constructing portable, rapidly deployable portal monitors to detect gamma radiation in passing vehicles. In this paper, we introduce a stereo system for precise, real-time, roadside tracking of vehicles in multi-lane roadways. The stereo system reports 3-D vehicle tracks to a gamma imager, which uses the track information to compensate for vehicle motion and produce detectable gamma peaks. We first describe the algorithmic components of the system, including calibration, camera motion correction, point tracking, stereo correspondence, and vehicle clustering. We also describe the physical system configuration. We then report results from tracking two different vehicles at speeds ranging from 5 to 20 MPH. We report on the accuracy of the system, where we observed a worst-case tracking error of 31 cm and an average tracking error of 9 cm in the direction of motion from a viewing distance of approximately 15 m.

Original languageEnglish
Title of host publication2011 Future of Instrumentation International Workshop, FIIW 2011 - Proceedings
Pages114-117
Number of pages4
DOIs
StatePublished - 2011
Event2011 Future of Instrumentation International Workshop, FIIW 2011 - Oak Ridge, TN, United States
Duration: Nov 7 2011Nov 8 2011

Publication series

Name2011 Future of Instrumentation International Workshop, FIIW 2011 - Proceedings

Conference

Conference2011 Future of Instrumentation International Workshop, FIIW 2011
Country/TerritoryUnited States
CityOak Ridge, TN
Period11/7/1111/8/11

Fingerprint

Dive into the research topics of 'Accuracy of a stereo vision system for portable roadside portal monitoring'. Together they form a unique fingerprint.

Cite this