Vehicle sparse recognition via class dictionary learning

Abstract

As the main body of modern traffic, transport vehicle is the focus of intelligent transportation systems. For three typical vehicles (including the automobile, motorcycle and bicycle), this paper proposes a new transport vehicle recognition system via class dictionary learning. For solving problems in the traditional transport vehicle recognition under sparse recognition framework, our method use SURF feature and class dictionary learning as the core. By the experiment with heterogeneous database, the effectiveness and feasibility of this method has been verified.