Research on Automatic Incident Detection Algorithm Based on Fusion of Freeway Mainline Information and Toll Collection information

Abstract

Incident management system is an important part of intelligent transportation system (ITS), and incident detection is the core of freeway incident management system. It is very important to freeway incident management system that incident is detected quickly and exactly. In order to realize automatic incident detection (AID), usually a set of vehicle detector need to be set every 500-2000 meters on the freeway. But now, in many countries, the factual detectors on the freeways are few, which can't meet the above needs. Most of freeway toll collection systems in many countries adopt close toll collection fashion. The establishment of computer toll collection system is perfect, and plenty of traffic flow information can be obtained from the toll collection system. This paper first brings forward the AID algorithm based on the mainline detector information and ramp toll collection information, which utilizes information fusion to realize freeway automatic incident detection and consequently improves the performance of incident detection algorithm. The algorithm can decrease interference to traffic flow when the detectors are being set. The paper studies the four freeway automatic incident detection algorithms whose input data are the mainline information, toll collection information and both information fused as data-layer fusion and character-layer fusion, respectively. As the result of simulation is shown, the performances of the freeway AID algorithm based on information fusion are better than those of the AID algorithms based on mainline information or the AID algorithms based on ramp toll collection information. And the performances of the freeway AID algorithm based on character-layer information fusion are better than those of the AID algorithms based on data-layer information fusion. The detect rate (DR) of the algorithm based on character layer information fusion is 93% and its false alarm rate (FAR) is 0.67%. The AID algorithm has a good potential application in practice.