CAR RECOGNITION BASED ON BACK LIGHTS AND REAR VIEW FEATURES

摘要

In modern societies, due to the high crime rates and high traffic accidents the feeling of insecurity and threat is increasing. The need for the establishment of defence and prevention mechanisms has encouraged studies to develop automatic recognition systems, which should, for example, have the ability to recognize vehicles at a certain distance. This paper introduces two car recognition methods, both relying on the analysis of car external features. The first evidences the shape of the car's rear view, exploring its dimensions and edges, while the second considers features computed from the car back lights, notably their orientation, eccentricity, position, angle to the car license plate and shape contour. Both methods are combined in the proposed automatic car recognition system. Experimental results indicate that car back lights and rear view features can be successfully exploited for automatic car recognition, showing a correct recognition rate of 89% on a database acquired at the IST parking lot.