VEHICLE CLASSIFICATION BY PARAMETRIC IDENTIFICATION OF THE MEASURED SIGNALS

Abstract: Traffic control requires measuring of road traffic parameters and vehicle parameters (velocity, length, number of vehicle, time distance between vehicles, etc.). One of very important parameters is the class of a vehicle. Up to date, videocamera systems, system with loop sensors, etc., have been used for classification. Unfortunately, the vehicle pattern were very complicated (picture from camera, time signal from loop, ...). Classification depends on the comparison of an actual signal to the reference pattern. It was very time-consuming and needed remembering the large reference pattern. This paper deals with a new method of vehicle classification. We have proposed the conversion of the measured signal into a vector of numerical parameters (only a few). The vehicle classification is being made by the comparison of such a very short vector with the vectors containing the values of parameters typical for the chosen class of a vehicle. Tests have been made on streets of Cracow (Poland). The classification efficiency for four predetermined vehicle classes is between 77% and 95%.