Vehicle Make and Model Recognizing Simultaneously by HVM Features

In this paper, a method to recognize vehicle makes and models simultaneously from a huge amount of surveillance images is proposed based on the vehicle frontal texture of air inlets. The moment features and de-correlation are utilized in recognition. The inlets of different model vehicle occupy different features even if they are belonged to the same make. In this paper, the inlet texture and moment are computed and these features are utilized to calculate the likelihood among different makes and models. To avoid the influence of correlation, de-correlation technique is utilized to calculate the difference among various makes and models. This method can recognize different makes and models of vehicles even when the makes are of just the same vehicles. In this case, the logos of vehicles are the same, but the inlet textures are different. The experiment verified the perfect performance of this method proposed in this paper when recognizing the makes and models simultaneously.