Vehicle make & model identification using Scale Invariant Transforms

Vehicle Make and Model Recognition (MMR) techniques provide an important functional enhancement to automatic vehicle identification systems that have traditionally been based solely on Automatic Number Plate Recognition (ANPR). A robust vehicle MMR technique should be capable of recognizing vehicle makes and models under varying environment and capture conditions. Most existing vehicle recognition systems use feature based approaches to first identify distinct features of different makes and models which are then matched under controlled criteria. Scale Invariant Feature Transforms have been proven to be capable of idetitifying invariant features which are scale, rotation invariant and are robust to affine distortions, noise, illumination and change in 3D viewpoint. It has recently been used with much success in image retrieval systems. In this paper we propose the use of a modified version of the popular SIFT based image matching approach to car MMR. We provide experimental results to prove its effectiveness in car MMR. We show that it is capable of identifying car make and model under image scale, blur variations and within practical limits of rotation, translation and occlusion.