Study On Vehicle License Plate Recognition Algorithm

Automatic recognition of vehicle license plate is one of the significant research subjects for applying computer vision, image processing and pattern recognition to the intelligent transportation and an important link in putting intelligent traffic management into practice. A typical automatic recognition system of vehicle license plate is composed of image capture, license plate location, character segmentation and character recognition etc.Algorithms of the main part of the system, including license plate location, character segmentation and character recognition, are respectively provided and realized.This paper researches on the algorithms of vehicle license plate location. Connected area analysis and horizontal cluster are used to locate the backup area of the plate in which the fuzzy decision is used to position the real area. This algorithm is excellent in accuracy and completely applicable to the complex traffic scenes.In the algorithm of the character segmentation homomorphic filtering is used to eliminate the influence from nonlinear illuminations, the least square method is used to update the inclination of the plate and deformable template is used to segment characters. It has an advantage of high percentage of segmentation.The least square support vector machine based on the experiential risk minimization of statistical learning theory is used to recognize characters in the character recognition and applicable to the small-sample machine learning. The paper makes use of character barycenter and wavelet to extract character feature, two-grid method to optimize parameters, and multiclass classification to recognize characters, which obtains excellent results in the practice. It is proved to be a higher accuracy, faster and elegant way of recognition, and satisfy the requirements of practical application.