Research On License Plate Location And Character Segmentation In License Plate Recognition System

With the rapid development of the social-economic as well as modern urbanizationprocess faster and faster, it is more and more demand for the number of cars, which broughtheavy pressure to the urban transport industry. So it is very much needed to establish of aintelligent transportation system. Automatic license plate recognition system in which occupyan important position,the study of it has practical application value. This paper will doin-depth research work on the image preprocessing, license plate location and charactersegmentation of the license plate recognition system,.In the image pre-processing section, first, the acquisition of a color image is convertedinto gray image type; For the actual brightness and contrast of the image, it select histogramequalization corrected image gradation or gray scale stretching method to improve the imagequality, and then use the median filtering method values to smooth image and to get a clearerimage; Finally,for binary image, it use improved morphological edge detection operator todetect the edge; Compared with the commonly used edge detection operator, the continuity ofimage edge is better by this method and provide the basis of the edge image for follow-up license plate location.In the design of license plate location, it put forward a method based on mathematicalmorphology and projection. First, mathematical morphology operations are used to locate theinitial license plate area; As the area is obtained based on relatively license plate geometricfeatures, so it needed to add the verification of the license plate area to get the correct resultsof the initial positioning; After it, in order to obtain the precise positioning, it use differentialprojection to get the horizontal and vertical directions of edges This approach reduces thepositioning error and effectively exclude the interference of the background of the image toaccurately posit the license plate.In character segmentation part, it study image binary of the license plate; Then,correctthe presence of tilt license and remove the borders and separators; After it, use mathematicalmorphology processing to intensive character binary image, then based on the charactervertical projection method and character boundaries to separate characters; The last, it achievecharacter size normalization and export standard character subgraph. The experimental resultsshow that the method is more effective to segment the license plate characters.