Application of Filtering Techniques for Incident Detection

Abstract

Traffic data, essential for the detection of freeway incidents, are often corrupted by impulsive noise and short‐term traffic inhomogeneities that may impair detection performance. Rigorous data filtering can reduce the undesirable noise and enhance the incident signal. A new incident detection algorithm is developed that employs short‐term time averaging (low‐pass filter) to reduce the adverse effects of short‐term traffic fluctuations and impulsive noise in the detection process. The detection algorithm traces the filtered spatial occupancy difference between adjacent detector stations through time, and detects an incident when this difference changes significantly in a short time period. The new algorithm was evaluated with data from a congested freeway in the Minneapolis‐St. Paul metropolitan area and compared against major existing algorithms. Test results indicate the ability of the Minnesota algorithm to significantly alleviate the false‐alarm problem, while preserving high detection performance, as compared to existing algorithms.