Non-linear compensation of vehicle signatures captured from electromagnetic sensors with application to vehicle re-identification

The distribution of travel times over a link in a transportation network can be estimated by observing the actual travel times of individual vehicles traversing the link and creating a histogram of the observations. The fundamental building block of travel time estimation is the re-idenfication of vehicles. A variety of techniques can be used for re-identification, but this paper focuses on electromagnetic signatures. The travel time distribution can be sampled using time-stamped signatures captured from vehicles along with an algorithm for signature matching. The method works well for free flowing traffic on a limited access highway, but to apply it on arterials it is necessary to compensate for vehicle acceleration over the sensors, which is manifested as a nonlinear scaling of the time axis in the captured signatures. In this paper we show how to estimate the acceleration from signatures captured from a loop trap and how to compensate for the effect of both varying velocities and acceleration. The results are summarized in terms of receiver operating characteristics to demonstrate the performance improvements associated with velocity and acceleration compensation. When applied to signature matching over a relatively long link it is shown that acceleration compensation offers significant improvement in matching performance.