Research On Technology Of Licence Plate Recognition Based On RBF Neural Network

License plate recognition is one of the core technologies of intelligent trafficsystem, and it is the integration of computer technology, image processing,pattern recognition and many other subjects. Thereupon, it is an importantresearch topic in the field of transport in recent years. At present, the licenseplate character recognition accuracy and speed has become a research focus.In this paper, the design of RBF neural network-based license platerecognition system includes: the image pre-processing, license plate location,characters segmentation, character recognition, etc. The system solves theproblem of license plate recognition.In image preprocessing stage, the image is treated by grayscale processing,median filtering, gray stretch and binarization processing. In the above process,OTSU method of image binarization processing is used to achieve the ideal effectand remove the noise of the image. Preprocessing stage lays the foundation forthe license plate location and character segmentation.According to the problems of license plate localization, this paper putsforward and uses a method based on mathematical morphological to achieve theprecise location of image after boundary detection process. Then the located plateimage is treated by tilt correction and frame processing to ensure its quality.Describing some basic features of license plate characters, this paper uses aimproved horizontal projection method to achieve characters segmentationaccording their features.After describing the basic structure and the learning process of the RadialBasis Function neural network and, this paper designs and trains a RBF networkto achieve the recognition of segmented characters.Finally, after extensive experiments, it is proved that using license plate recognition method based on RBF neural network can achieve good recognitionresults and a high recognition rate.The mathod has a high practical value.