Vehicle recognition based on Gabor and Log-Gabor transforms

Image-based vehicle recognition is usually addressed as a supervised classification problem. Here, vehicle recognition is performed using two different transforms such as Gabor and Log-Gabor. First the images are convolved with Gabor and Log-Gabor filter with different scales and orientations. Then mean and standard deviation are computed for all the filtered images. These features are fed to SVM classifier for further training and classification. The experimentation is carried out in GTI database. From the experiments it is revealed that the Log-Gabor transform significantly outperforms the standard Gabor transform for image-based vehicle recognition.