Research And Implementation Of License Plate Recognition System Base On Image

License plate recognition has very important value in intelligent transportation system,which is crucial to remember and intelligent transportation systems. Many experts andscholars have started the in-depth research on license plate recognition, and have manyoutstanding contributions to the theory and algorithms. License Plate Recognition combinesimage processing, pattern recognition etc. that can be widely used in parking management,high-speed toll management, which can save many man and material resources. This paperfocuses on three essential aspects in license plate recognition systems involved positioning,segmentation and registration. We implement a simple license plate recognition system underC language and OPECNV library. We summarize the whole work are the following:First, the paper describes the background and significance of the research. By analyzingthe research background, significance, this paper illustrates the research and implementationof an image-based license plate recognition system has a certain necessity.Second, the paper researches some license plate location method. License plate locationis the premise of segmentation and recognition. Therefore, the accuracy of license platepositioning directly affects the accuracy of license plate segmentation and license platerecognition. This article describes four common positioning method: Edge-based approach,morphology-based approach, color-based approach and projection-based approach. We founda single license plate location positioning accuracy is not able to meet the needs, so wefinalize the license plate location method based color division and projection. In the algorithm,we found that the fusion of these two methods could be well positioned the license plate.Third, the paper researches some license plate segmentation methods. This paperdescribes two simple license plate segmentation methods: connected-domain method and thevertical-projection method. By analyzing these two methods, we found that thevertical-projection algorithm has a good result on speed and effect.Fourth, the paper studies some license plate recognition methods. This article describescommonly used three methods in license plate recognition: neural networks based method,feature-based statistical method and template matching method. By principle analysis and experimental, we find the method based on neural network has better recognition accuracyand anti-jamming capability, which is good for high accuracy license plate recognitionsystem.Fifth, through the study of plate location, segmentation and recognition methods, thepaper implements a simple image-based license plate recognition system. In recognitionsystem, license plate location is using a color-based method and projection-based method,License plate segmentation method is using a simple vertical projection based, and licenseplate recognition is using the BP neural network method. In experimental tests, theimplemented system in this paper can good at license plate positioning, segmentation andrecognition.