Research On Recognition Technology For Vehicle License Plate

Vehicle license plate recognition technology can be applied not only in the field oftransportation administration, vehicle management but also applicable at other publicplaces like toll stations, parking lots. Thanks for the rapid development of imagerecognition technology, it penetrate tremendously into wide range of applications at allassociated field today bring great convenience to people’s lives and socialmanagement. Therefore, further study of car license plate recognition technology isstill necessary and could greatly improve the development of the managementefficiency and capacity meanwhile contributes its significant academic value.Based on the comprehension the crucial technology of identification license plate andrelevant key technologies, this dissertation focus on deepen research of imagepre-processing, automobile license plate image orientation, image segmentation andimage recognition and other aspects as following:1. For the study of image preprocessing techniques, the author deeply studied thegrayscale of the image, noise reduction, image enhancement and binarization.Simulation results showed that the image quality of license plate was improved afterimage preprocessing.2. For the study of license plate location, the research is based on two methods:mathematical morphology and texture feature with comparing the simulation effect ofthese two methods. The results prove the validity and accuracy of precise positioningmethod.3. In the study of character segmentation technology of license plate, the charactersegmentation methods is studied respectively based on Unicom area and projectionmethod, and illustrated separately their applicability of the two methods. Throughcomparing the segmentation result, the author found out that the projectionorientation could be better. 4. In the study of character recognition technology of license plate, the authorproposed the character recognition methods which is based on template matchingand BP neural network with integrating the better methods of Image preprocessing,positioning and segmentation. The author used MATLAB simulation to contrast therecognition effect of33images of random car license plate. The results showed thatthe recognition effect based on BP neural network was more accurate and reliable.