Research On Chinese Character Image Recognition Machine

Aiming at the script Chinese character images and the VLP Chinese character images, the stabilization characteristics, the holistic characteristics and resolvability are researched in this thesis to construct the Chinese character image recognition machine. The coarse classification, characteristics pick-up, judgement and feedback are used in the machine to increase the recognition adaptability and accuracy.The following are the major contents in the thesis:(1) A apery Chinese character image intelligent recognition machine. The apery (expert) experience, pattern recognition and the process of intelligent control judgement and feedback are researched. A apery Chinese characters image intelligent recognition machine is presented. A paralleling structure of decision-making level, pattern recognition level and feedback level is adopted. Decision-making level chooses the best recognizing strategies and parameters, Pattern recognition level uses multi-classifier results fusion, Feedback level can feed back to revise the recognizing methods and parameters depending on the results of reliability judgement. This structure can increase the recognition rate.(2) A Chinese character image coding recognition machine. The same and different characteristics of many Chinese character image coding recognition methods are researched. A Chinese character image coding recognition machine is presented. It can choose the steady characteristics of each coding recognition methods to make coarse classification, then, use system feedback recognition depending on different situation and memorize the recognition results to judge and queue the recognition effect of every coding recognition methods dynamically. It can increase the classification rate and recognition rate.(3) Depending on the interrelated literature datum and the former research results of my lab, the wavelet transform and other mathematic instruments are adopted;the VC++ and the MATLAB6.5 are used as the realization methods in this thesis. The recognition of various style script Chinese character images and the VLP Chinese character images depending on the above two Chinese character image recognition machine structures are accomplished successfully. The more recognition times, the more experience the machine gets, and the higher recognition rate.The research in this thesis shows that the above two Chinese character image recognition machines can make coarse classification, the best recognition strategies decision-making, multi-classifier results fusion recognition, reliability judgement and feedback revision depending on the recognition results. These can increase the recognition accuracy.