A Low Pass Filter for Incident Detectionby Yorgos J. Stephanedes, Univ of Minnesota, Minneapolis, United States,
Athanasios P. Chassiakos, Univ of Minnesota, Minneapolis, United States,
Abstract: A new algorithm for automatic incident detection was developed after observing that existing algorithm performance can improve if detector data are filtered before use. The new algorithm lies within the traditional framework of comparative algorithms, i.e., is based on simple comparisons extracted from loop detectors. The development focused on using filtered detector output, i.e., values averaged over short time periods, to avoid false alarms that are primarily due to short-term traffic inhomogeneities. The algorithm was tested against major algorithms reported in the literature. The evaluation included main features of the algorithm, i.e., detection rate, false alarm rate, and mean time-to-detect. Test results with incident and incident-free data along I-35W, a typical freeway in the Twin Citites Metropolitan Area, are presented via its operating characteristics curves (a plot of detection rate vs. false alarm rate).
Subject Headings: Algorithms | Filters | Probe instruments | Traffic accidents | Traffic analysis | Traffic surveys | Traffic signals | Comparative studies
Services: Buy this book/Buy this article
Return to search