Research On License Plate Location And Slant Correction As Well As Character Segmentation In Vehicle License Plate Recognition System

Vehicle license plate recognition system(LPRS) is the hardcore of the intelligent traffic system(ITS).It can be used in the fields such as electron charging,pass controlling,automobile stream supervising and so on.LPRS consists of four main modules which are license plate location,slant correction,character segmentation and character recognition.The former three modules have great effects on the last module.This paper has summarized a lot of materials and does research on license plate location,slant correction and character segmentation.In the stage of license plate location,color segmentation is applied firstly in HSV color space to extract color pair pixels according to the rule that character color and background color should be matched.Then,mathematics morphological operation is adopted in order to get some continuous lines and texture features of these lines are analyzed so as to filtrate disturbed lines.After that process the remained continuous lines are combined to locate the plate.Finally,the whole plate is got through expanding the plate area.In the stage of slant correction,a horizontal slant correction approach based on color pair pixels and principal component analysis is presented.Through principal component analysis of the color pair pixels,the slant angle of the plate is computed and the horizontal slant correction of the plate is completed.After that,vertical slant correction is also completed and the border of plate is obtained.In the stage of characters segmentation,pretreatment including binary and filtering is taken at first.Then,a method based on projection and pre-known knowledge is put forward which can also deal with those characters that are broken or combined together.Finally,a software platform is constructed by VC++ programming tool.This platform contains all the steps that mentioned above.Extensive experiments are conducted over a lot of real-world vehicle license plates.The result shows the algorithm in this paper is effective and prepares good conditions for the continued character recognition.