Graph Grammar Based License Plate Recognition

Automatic license plate recognition is still a challenging problem where severe imaging conditions must be handled. As compared with other formal tools, the graph grammar model is more suitable when dealing with the severe imaging conditions. In this paper we introduce a graph grammar based method of license plate recognition. In our approach, a level set based license location process is first employed to find candidate plate locations. Then a graph grammar is applied to each candidate location to recognize the plate. The grammar consists of 3 rules. The first two rules recognize individual characters. The third rule represents the license plate layout by a graph. The plate recognition is achieved by a graph matching process.