Research and Comparison of Several Vehicle Detection Methods in Urban Traffic Scenes

Urban traffic road environments are complex and we present a vehicle detection method based on symmetrical features for vehicle contour and license plate. First, the vertical symmetry axis of vehicle contour is detected by a horizontal scanning line, and the horizontal and vertical symmetry axes of the license plate are detected. Finally, the vehicle area is located according to the license plate symmetry axis. Several vehicle detection methods in urban traffic scenes, such as those based on symmetrical features, license plates, gabor features, and support vector machines (SVM), and Haar-like features and AdaBoost classifiers are compared in this paper. The theoretical analysis and experimental results show that the method of vehicle detection based on symmetrical features effectively eliminates environmental influences such as noise and road markings on vehicle area detection. The detection rate of this method is superior to three other methods and the detection accuracy is up to 94.2%.