Vehicle Model Recognition Based on SURF

摘要

At present, the recognition of vehicle's brands and model is more and more hot in the field of general intelligent transportation. This technology has been widely used in city traffic monitoring, parking management, highway tolls and some other fields. In this paper, a recognition algorithm and corresponding system is proposed based on SURF operator. Firstly, by the license plate positioning, we define a novel Region of Interest (ROI) for the front view of cars. Then the ROI is segmented and corresponding SURF descriptor in ROI is extracted. Finally we use SURF descriptor to complete recognition. Numerical experiments show that this method has the recognition rate of 95%. The recognition ability is reliable. This algorithm can deal with the interference of different image scale, and also could not be affected by the illumination variation and weather conditions changing.