This paper presents a new method to combine the traffic incident detection data from multiple sensors using Dempster-Shafer's algorithm, an evidential reasoning based data fusion procedure. The aim is to solve traffic congestion problems associated with roads of limited capacity. In order to generate a timely response to traffic congestion, quick and accurate incident detection is desirable so that a tactical picture can be made available for situation assessment. In this paper, the focus is to assign consistent and intuitive weights to incoming incident reports in order to accurately discriminate between true incident and pseudo (false) incident. The concept of Dempster-Shafer's algorithm was used to combine sensor reports and reduce the false alarm rate in incident detection for providing accurate assessment to various traffic situations. The concept is presented with example scenarios and its performance is compared to conventional incident detection algorithms. (A*)
Abstract