Automatic vehicle classification using fast neural network and classical neural network for traffic monitoring

Abstract

This paper introduces an automatic vehicle classification for traffic monitoring using image processing. In this technique the fast neural network (FNN) as a primary classifier and then the classical neural network (CNN) as a final classifier are applied to achieve high classification performance. The FNN gains a useful correlation between the input and the weighted neurons using a multilayer perceptron to provide detection with a high level of accuracy. The Fourier transform is used to speed up the procedure. In the CNN, a lighting normalization method is employed to reduce the effect of variations in illumination. The combination of the FNN and CNN is used to verify and classify the vehicle regions. False detection is added to the training procedure using a bootstrap algorithm to get nonvehicle images. Experimental results demonstrate that the proposed system performs accurately with a low false positive rate in both simple and complex scenarios in detecting vehicles in comparison with previous vehicle classification systems.