Vehicle Re-Identification With Dynamic Time Windows for Vehicle Passage Time Estimation

A simple method for vehicle re-identification to generate vehicle passage times with loop data is developed. The method departs from other existing methods for vehicle passage time estimation: 1) It handles vehicle signatures one at a time and evaluates each vehicle observed only once. 2) The commonly used prespecified time window is replaced by a dynamic list of vehicles to be matched. 3) Vehicle matching is based on a combined estimation model that integrates spot traffic data with spatial vehicle data. The performance of the algorithm was tested with field data. Furthermore, to examine the effect of some of the assumptions on the performance of the algorithm, we compared the result with that obtained from an offline optimization model based on a spatial constraint that considers as many vehicles as possible for matching. The proposed method is particularly suitable for real-time applications since it can be easily implemented with little calibration effort and is computationally efficient.