Car License Plate Recognition Technology Research

Mobile License Plate Recognition (LPR) is an important part of Intelligent Transportation System, and is also an attractive research in image-processing and pattern recognition areas in recent years. So it has important theory value and application value to make the research of LPR..Based on the gallery delivered by one gas station, this paper do some research on every step of LPR. It also put forward or advanced some methods. With applying these methods, good results have got. The main studies are as following:The first step is to locate License Plate in the picture. The paper proposes a simple method which based on texture filter and median filter. The correct ratio of location is 98.8% based on all the test pictures.In the stage of plate image preprocessing, the main work is to adjust the tilt License Plate and binarization. To correct a tilt plate, a method was put forward to obtain the horizontal or vertical incline angle by Hough method calculating the edge of the plate. Binarization is also significant in the course of pre -treatment The method used in locating License Plate is also used to get the accurate characters area. Then the clustering method is applied to get the background color of the plate. A different threshold value is used according to different background color.In the step of character segmentation, this paper presents an adaptable method which can get the exact borderlines of the character. Normalization doesn't use the interpolation method, but apply a method using magnifying picture and dwindling the zoomed picture. Then the standard picture is got.Character Recognition is mainly based on template matching. Some letter character and figure character may be confused by template matching. Then the neural network is to distinguish them. BP method is used in the paper.The correct ratio of location is 98.8% based on all the test pictures. The correct ratio of Character Recognition is about 96% based on the test pictures of this step.