n South African Computer Journal - Empirical results of an image-based automated vehicle identification system : research article
|Article Title||Empirical results of an image-based automated vehicle identification system : research article|
|© Publisher:||South African Computer Society (SAICSIT)|
|Journal||South African Computer Journal|
|Author||L. Coetzee and E.C. Botha|
|Publication Date||Dec 2003|
|Pages||33 - 39|
|Keyword(s)||Correlation classifier, Frequency domain analysis, I.4, I.5, Linear classifier, Metallurgical fingerprints, Nearest neighbour classifier, Neural net classifier, Principal component analysis and Vehicle identification|
This paper describes the image analysis and processing of an automatic vehicle identification system. The envisioned system will identify vehicles based on video or still images of unique identifying marks, namely metallurgical fingerprints, as proposed by De Kock in his South African patent . Two candidate image analysis algorithms for automatic vehicle identification are selected. The selection criteria and the process for identification of these algorithms are described. The algorithms were tested on a dataset of real-world vehicle images. The experimental approach and the results obtained from applying these algorithms to the automatic vehicle identification problem form the main theme of our paper. Based on the results, the recommended system uses a nearest-neighbour classifier with features as extracted with a wedge-ring detector from the frequency domain of the image on a histogram preprocessed dataset. Finally we present excellent verification results using the proposed approach on separate verification dataset.
Article metrics loading...