Section-Related Measures of Traffic System Performance: Prototype Field Implementation

In this project (MOU 336),an initial phase of a field implementation was accomplished of the results of a previous research project (MOU 224),in which a vehicle reidentification algorithm based on loop signature analysis was developed using freeway traffic data.This algorithm was extended to non-freeway cases, initially using a section of 2-lane major arterial in cooperation with the City of Irvine,California.The technique was enhanced to address problems such as "irregularities " in vehicle signatures associated with trucks,tail-gating vehicles and erroneous counting of vehicles,with the objective of obtaining 100% correct counts at each station.The enhanced algorithm was also applied to a major specially instrumented signalized intersection in Irvine,California to demonstrate acquisition of data for real-time congestion monitoring,incident detection and level of service measurement.The initial application was for through vehicles on one approach.In order to achieve more reliable vehicle reidentification results,additional routines for vehicle movement filtering at the downstream station were applied.Reidentification results based on an initial dataset showed an encouraging matching result of 84.07%overall,for individual vehicles.Speed estimation from a single loop signature was one of the applications investigated in detail. For several study sites,the vehicle reidentification matching rates,using speed estimated from a single loop,were only slightly lower than for double loops.In another detailed application (using freeway data collected in previous PATH project MOU 224),vehicle classification using a Backpropagation Neural Network showed an 80 %classification rate overall for all vehicle types.Heuristic approaches to vehicle classification also demonstrated good results.