License Plate Localization in ALPR Systems

In this paper, an approach to vehicle license plate localization is described. The algorithm starts by isolating objects, in the image, that can be possible candidates of characters in a license plate. It then uses distances between objects and their relative positions to identify possible groupings (series) of characters that could belong to a license plate. The algorithm then uses a novel character recognition approach to recognize the identity of each character in the series. Finally, the relative positions of series are used to localize the license plate. The algorithm was successfully applied to images containing Lebanese license plates.