Car type recognition in highways based on wavelet and contourlet feature extraction

Recently many works focus on the vehicle type recognition because it is important in security and authentication systems. Computational complexity and low recognition rate especially when the system has to recognize among a large number of vehicles, are two major problems in vehicle type recognition. In recent years wavelet and contourlet transform have been applied in the recognition tasks successfully. In this paper we proposed a method for recognizing vehicle type in different lighting conditions. We used wavelet and contourlet as tools for feature extraction. These features are powerful and robust to illumination and scale variation. We reduced the dimension of feature vector by resizing the wavelet and contourlet subbands and then applied normalization on those coefficients. Our method is robust to a few variations in vehicle frontal view angels and distance to camera. The experimental results showed 97.35% true recognition rate for 14 classes of cars which is a significant increase for vehicle type recognition.