Abstract:
License plate recognition has been a popular research topic for a long time. Currently, there are a lot of commercial software packages available but very little effort has been done to recognize province names in Thai license plates. A reason that province name recognition is not popular may be due to low quality of the input of character images. It is very difficult to recognize the province name in the plate by separately recognizing each character image because the boundary of each character image is not sharp. Therefore, this thesis avoids this problem by viewing a province name as one single image, and not recognizing individually each character image. Then, the probabilistic models, which have been shown to handle sequence data effectively, are applied to the recognition of province names from the image. The experiments are conducted to evaluate the proposed method by comparing conditional random fields with a linear-chain hidden markov model, and the results show that conditional random fields provide accuracy higher than the hidden markov model.