Vehicle model recognition based on using image processing and wavelet analysis

Abstract

Today, using intelligent transport systems such as identifying vehicle type for monitoring traffic in urban areas can advantage a lot. In this paper, a new method has been presented to detect vehicle type with only one reference image per class. Our algorithm is based on searching mean gradients and analysing these changes by Daubechies wavelet transformation. Firstly, a feature vector is obtained based on the car boundary changes of the side view image in proposed system. This vector is then analyzed by Daubechies wavelet transformation and three statistical criteria; variance, norm-1 and norm-2 are extracted from wavelet coefficients. Finally, a similarity factor is defined to detect the same type of vehicles. The proposed algorithm is resistant against car edge negligible changes and the experimental results indicate the high performance of system in detecting the vehicle types.