Vehicle Re-Identification With Several Magnetic Sensors

Abstract

Vehicle re-identification gives access to two essential data for traffic management: travel times and origin-destination matrices. This paper aims to evaluate the performances of a vehicle re-identification method based on Euclidean distances between vehicle “magnetic signatures” measured with several three-axis magnetic sensors. These performances are compared to the ones achieved by a single sensor and the three-dimensional (3D) Dynamic Time Warping algorithm. Moreover, the effects of a change in the vehicle orientation or in its lateral position are studied. On a data base of signatures from 25 different vehicles, the best results are obtained with the 3D Euclidean distance: 100% of the pairs of signatures are correctly re-identified without any false alarm if the vehicle keeps the same orientation and 90% if the vehicle orientation changes. The influence of the vehicle orientation to the magnetic North on the signature is therefore limited. However, the performances fall when the vehicle lateral position shifts.