A neural network based system for vehicle classification

In this paper we describe a system used to recognize and classify vehicles traveling along a roadway. The system was developed using standard video cameras placed above and along the side of the roadway rather than requiring specialized cameras placed at specific locations, such as attached to an overpass. The classification of vehicles is performed by a synthesis of multiple sets of features. Four sets of features are currently being used. They are, the reduced wavelet transform of the vehicle image; a similar wavelet transform of the wheel section of the vehicle; the edge outline of the vehicle; and a set of normalized vehicle attributes; namely, length, width, and height. In order to use any such feature set, it is necessary that the image be normalized and thus that the features be invariant with respect to size, and position. The wavelet transforms and features are used as inputs to a neural network. The output of this network is the vehicle's category.