Study Of Algorithms On License Plate Location And Characters Segmentation

In Intelligent Transportation System(ITS), the recognition of the license plate by computer has very important significance in the society, such as toll gates in highway, vehicle survey, monitor in parking area and management system. This paper give further investigation to the license plate location in ITS and character segmentation, include the following content concretely:Researching the algorithm of license plate location. First, pretreating the images, including gray the RGB image, gray stretch and bring forward the segmentation method of vertical texture characteristics of characters based on edge detection. This method is forming obvious edge by frame characters and background of license plate to carry out the location, and accomplish segmentation of all kinds of license plate by geometry characteristics combine pixel distribution characteristics in selection fields. The experimental results shows high recognition rate and fast speed for license plate.Researching the method of license plate pretreatment. First, after analysing the present binary algorithms, combine the characters of global threshold and partial threshold binary algorithms, this paper bring forward a improved binary algorithm based on CASDA and Bernsen, accomplish the quick effective binary to license plate.This article brings forward a method based on edge detection to correct slant plate while Hough is difficult to get the right angel in different color image.Researching the segmentation algorithm of license plate. According to the rules of characters and geometrical features of the characteristics, an improved algorithm for character segmentation is proposed which is based on projection. Firstly, get rid of the disturbance of edge by using line scanning and projection. And then bring forward a solution to character conglutination and rupture which are difficulties by vertical projection. Introduce the concept of narrow character and wide character to classify the character block field. Finally the characters of the license plate are segmented accurately. The experiment shows that this method is good for character segmentation in a short time. The algorithm we adopted is not restricted by the location of the plate in the image, and it has strong anti-jamming ability to lean, distortion, pollution and blur. The algorithm is also adaptable to the varieties of circumstance light and image contrast.