A Number Plate Recognition System for Abnormal Data

In view of complex background the license plate image which photographs under the different illumination condition, in this research, we proposes one in administrative levels fast localization method. This method use license plate geometry characteristic, textural property as well as color characteristic. Firstly we take a much more quicker interleaved method of sampling according to the gray texture feature of the license plate. Secondly, to remove the most of the background unrelated according to the geometry feature of the license plate. Thirdly, to calculate the gray statistics of the candidate area using the method of consecutive scanning and the color statistics, so that we can get the accurate center. Then the trump card expands according to the central region for may contain the entire license plate´s subgraph.