Vehicle Type Classification Using Convolutional Neural Network

Automation of vehicles' classification and recognition is one of the most important challenges in contemporary road safety and intelligent transportation system. The development of image processing, pattern recognition and deep learning technologies has overcome many obstacles to achieve this aim. In this paper, we present a vehicle type classification system based on deep learning technology. This system is constituted of two steps. In the first step, we apply data augmentation to attenuate the imbalanced dataset problem. In the second step, we build a convolutional neural network (CNN) model with different architectures using parameters that are learned from the training dataset. This system is part of a integrated application that will enable automated traffic signal management based on vehicle type automatic detection.