Salient features based on visual attention for multi-view vehicle classification

The continuous rise in the amount of vehicles in circulation brings an increasing need for automatically and efficiently recognizing vehicle categories for multiple applications such as optimizing available parking spaces, balancing ferry load, planning infrastructure and managing traffic, or servicing vehicles. This paper describes the design and implementation of a vehicle classification system using a set of images collected from 6 views. The proposed computational system combines human visual attention mechanisms to identify a set of salient discriminative features and a series of binary support vector machines to achieve fast automated classification. An average classification rate of 96% is achieved for 3 vehicle categories. An improvement to 99.13% is achieved by using additional measurement on the width and height of the vehicles.