Vehicle Type Classification from Surveillance Videos on Urban Roads

Abstract

In this paper, a novel classification scheme has been proposed for real time vehicle type classification from surveillance videos on urban roads. Three kinds of vehicle types, i.e., Small cars, large cars, and motorbikes, are classified for the later retrieval. This system is performed in various outdoor illumination and weather conditions. The average precision and recall rates of vehicle type classification are more than 93.82% and 88%, respectively. Moreover, the performance of the proposed method is up to 25 frames per seconds.