SMOOTHING ALGORITHMS FOR INCIDENT DETECTION

The majority of automatic incident detection algorithms aim to identify traffic incident patterns but do not adequately investigate possible similarities in patterns observed under incident-free conditions. A classification of major traffic disturbances on freeways is presented. On the basis of this classification, an incident detection logic is developed with the traffic features that result in the best distinction between an incident and other disturbances. The new logic, DELOS (Detection Logic with Smoothing), employs smoothed detector occupancy measurements to signal an incident when significant temporal changes of the smoothed occupancy occur. Three types of smoothers--average, statistical median, and exponential--are considered, leading to corresponding algorithms. The structure of the proposed algorithms is presented and compared with previous algorithms. Comparative evaluation of test results with rush-hour traffic and incident data from I-35W in Minneapolis reveals the improved performance of the proposed method.