Research On License Plate Recognition Technology In Large Angle

With the rapid increase of car ownership, the car has become the main means of transportation, license plate recognition technology plays an increasingly important role in public transport and daily life. The existing license plate recognition technology can meet the requirements of the general situation, but in the large angle of the license plate tilt occasions, there are still many deficiencies. So this thesis mainly focuses on the license plate recognition technology in large angle to do some related research.In this thesis, we study the typical license plate recognition technology and analyze their characteristics, and a series of studies are carried out based on this simulation. In license plate location, the gradient image is extracted by horizontal difference operator,and the initial location of license plate is realized by morphological operation, and the accurate location of license plate is achieved by projection analysis. In character segmentation, fast correction is achieved by extracting the lower edge of the character and fitting a straight line, which is more rapid and robust compared to the traditional method.For character recognition, the character template is weighted first, and then the left contour of the character to be recognized is extracted, and then a algorithm for template matching of image division block is used to reduce the amount of computation.The experiment results show that the proposed algorithm can locate the license plate quickly and accurately, and realize the correction and recognition of the license plate in large angle by the experiment of 382 car images from different scenes, the total recognition time is less than 0.3s, and applicable to complex background.