| License plate recognition (LPR) is the technique to draw the character information from the vehicle image and confirm the identity of the vehicle through using computer vision, image manipulation, pattern recognition, and so on. LPR is a important study field in the ITS, and it has great apply value in the common security, traffic management, flow observation, electronic charge system and relate military department.LPR is divided into plate position, character segmentation, and character recognition. Plate position is a difficult problem: plate area has small proportion in the plate image, the color, size and position of the plate is uncertain, and the position arithmetic should overcome the influence of different illumination and complex background, moreover it must be accurate and fast. The thesis studies a new method of edge detection combined with scan lines which solves the problem beautifully.In the process of getting image, due to the change of the angle between the camera and the plate, the vehicle image usually slants, that leads to the distortion of the plate and the characters, and brings great influence to the character segmentation and character recognition. So a new method of bilinear interpolation for correction is proposed which has the advantages of rapidness and effect.Character segmentation is the key step of the LPR and is the base of character recognition. A segmentation method of vertical projection combined with templates matching is brought forward and it solves those problems such as non-integrity, position shift, character abrasion, character distortion, and so on.Character recognition is the core of the LPR. The thesis discusses the traditional method of pattern recognition and the application of the ANN in the Character recognition. After that a method based on BP-NN is introduced, and it is proved to have a high rate of recognition by experiments. |
