A Study on Contour Feature Algorithm for Vehicle Type Recognition

Abstract

Vehicle type automatic recognition is of great important today in Intelligent Transportation System. And neural network is often applied to recognize the vehicle type. However, the network can be very complex and therefore difficult to be trained. In order to cope with such issues, a new developed vehicle type recognition method based on contour feature is presented in this study. It is applied to obtain the vehicle type from the geometrical feature of the vehicle. This enables the implementation of the recognition system only in given geometrical size and simplifies the thinning recognize procedure. The contribution of this work is threefold: At first, a novel evolutionary methodology for extracting vehicle feature is presented. Secondly, a vehicle recognition algorithm consisting of four steps is demonstrated. Finally, the performance of the recognition system is evaluated by not only using static vehicle image but also using dynamic vehicle video.