Research On Technology Of License Plate Automatic Recognition

The technology of license plate automatic recognition is one of the important research fields in intelligent Transportation Systems (ITS). It has a wonderful future.License plate recognition (LPR) includes three important techniques which are the location of vehicle license, the segmentation of the characters of the vehicle license and the recognition of the character of the vehicle license. There are many ways of recognizing the characters of vehicle license, but the recognition of vehicle license always is a difficult question when the characters are blurred, inclined, broken, distorted and smutted.In this paper, the studies are focused on the algorithms for LP image preprocessing and recognition of characters of vehicle license.A vector median filtering algorithm for color image based on direction region distance measure is proposed in the paper. For traditional vector median filter, its computation cost rapidly increases and the speed of the algorithm is influenced when the size of the filtering window increases. In this paper, direction region distance measure is defined based on the region attribute of pixels and the window is divided two sub-windows. And then the filtering sub-window is confirmed by average color distance between sub-windows and the window. Finally, the vector median filtering algorithm is performed in the sub-window. The experiments carried out show that not only more details of color images are preserved but also the cost of computation of the algorithm is reduced and its speed is increased.A new algorithm for number and letter character recognition of vehicle license plate based on 3 levels template matching is proposed. Firstly, foreground matching measure and background matching measure are used in the first level template matching. If similar characters are gained, characteristic region template is used in the second level template matching. If the first level matching fails, hollow templates are applied in the second level template matching. If the second level matching fails, the third level matching is the hollow template matching based on Hausdorff distance. The experimental results show the effectiveness of the algorithm. It not only can identify the similar characters, but also offers robustness to the dirty, broken, distorted characters.