Inductive-Loop-Based Vehicle Signature Features Analysis and the Anonymous Vehicle Reidentification for Travel Time Estimation

Traffic planning or its management decisions are essentially offline by analyzing data. To improve the traffic performance, one analyses real time data collected from various sensors. The widely used sensor is certainly Inductive Loop Detector in many countries such as France. However it is quite recent that researchers interest in its non-binary use. Thus, the vehicle signatures are now analyzed in order to recover vehicle speed or to classify vehicle categories. The applications seem to be promising: travel time and O/D matrices estimation and vehicle tracking. The principle is to analyze the signal and to extract features that allow vehicle identification. These applications encounter mainly two problems: features selection (what are the reliable and independent features?) and limited data transmission capacity. The purpose of this paper is to present our approach to select independent features in order to use as less as possible features and as efficient as possible and the results of our work.