Efficient vehicle identification using MPEG-7 Color Layout Descriptor

Vehicle identification is an intersection research topic of Intelligent Transportation and pattern recognition which has been attracted much attentions. However, the matching precision of vehicle image and computing cost are hard to balance. This paper proposes a vehicle identifying scheme which combines extracting vehicle object based on Frame Difference (DF) method and image matching based on MPEG-7 Color Layout Descriptor (CLD) aiming at pursuing a trade-off solution. Experiments on real dataset from surveillance videos of Shanghai bureau of transportation management show that the proposed algorithm could boost image matching speed greatly while obtain fairly good performance as classic CLD method.