The aim of Chinese character recognition is to make the Chinese input more natural and convenient so that the computer could process Chinese information more easily. In practice, large volume of letters, newpapers,magaiznes need to be coverted into a coded representation of the input characters. That's what printed Chinese character recognition can do. And practical demands, such as business card recongnition, identification card recognition urges more achievements in printed Chinese character recognition. Firstly, this thesis gives an introduction of the history and the development trend of Chinese character recognition. The property and feautres of Chinese character are pointed out. Then I give a rise to the projects relate to this thesis. In chapter 2, the basic procedure of Chinese character recognition is introduced. Every step of recognition is elaborated in details. Especially, I analyze the hot research aspects of this field, for example, the combination of multi-classifier. In chapter 3, the common features of Chinese character recongnition are compared, such as edge detection feature, mesh feature, directional distance distribution feature. A conclusion is drawn that combination of multi features is an effective way to improve recognition rate. We present the source and the basic concept of wavelet theory in chapter 4 and 5. The strongpoints of multiresolution analysis are verified. And then we use it in practical system and propose the multiresotution recognition system of printed Chinese character with wavelet transform. This method give a new view of printed Chinese character recognition, and more exciting is that it can be used in other recognition problems. In chapter 6, the basic methods of context-based post-processing are introduced and a knowledge-based system for business cards understanding is brought forward. Experiment results indicate that this system meets the requirement of practical projects. In this thesis, we describe printed Chinese character recognition system in details and propose the multiresolution recognition method. This method is a new attempt in the field. The knowledge-based context post-processing method is also presented, which gain many praises from users. With the efforts of rescarhers, more and more OCR products will appear on the market and the Chinese information processing will become more convenient.
修改评论