Research On License Plate Detection And Recognition Technology

License plate detection and recognition (LPDR) technology, as one of the important research subjects of image processing and pattern recognition in the field of intelligent transportation, is widely used in the residential parking management, toll collection and traffic management. It has a great value of market and theory.LPDR mainly consists of three parts:license plate location, character segmentation and character recognition. The difficulties of license plate location lie in poor lighting condition, complex background, color drift and contaminated license plate. The existing algorithms of character segmentation can not split the rupturing or adherent characters. Because of the irregular strokes and incompleteness, the recognition rate of Chinese character is relatively row. On the basis of the analysis and summary of current existing LPDR technology, this paper proposes a complete technical solution of LPDR to alleviate the difficulties of existing algorithms. Main study work of this paper can be summarized as follows:1. License plate location. For poor lighting, it designs a custom edge detection operator to strengthen the texture. It integrates the color and texture feature of license plate to solve the color shifting and complex background interference. Finaly, the real plate is located according to the regional centers. Experiments prove that the rate of license plate location can reach97%.2. Character segmentation. Firstly, it corrects the license plate with PCA algorithm. It then uses the vertical projection to find the basic segmentation information, and decides the self-adaptive width of character by analyzing the peaks and valleys, with which it checks the fracture and adhesion of character from the middle peak. On the basis of completing the computation of peak probability, it selects the seven consecutive peaks with maximum probability as the segmentated characters. In complex cases the rate of character segmentation reaches92%, which means it solves the problem of character fracture and adhesions effectively.3Character recognition. It builds the standard Chinese character and half-word template library. Then it compares the template characters with the character to be identified by custom match operator that is able to recognise the irregular and incomplete character. For the English and digital characters, it uses three layers of BP network to recognize the characters. In the process of training the BP network, it updates weight and bias values according to gradient descent with adaptive learning rate. With more than95percent of recognition rate, they achieve the expected goal.