Formula Character Recognition And A BP Parallel Algorithm Classifier

With the improvement of the storage capability of computer, more and more documents, papers, articles are scanned into computers and saved in images. However, these images can not be reedited. Nowadays, the technology that converts document images into retrievable and editable forms is more and more concerned by researchers. Document image analysis (DIA) comes into being to do this job. Optical character recognition (OCR) is the core DIA dealing with either printed or handwritten document. Usually, there are many mathematical formulas in scientific documents. These formulas usually have Greek characters and other special symbols, and there often exist two-dimensional position relationships among the symbols of these formulas. At present, there is no OCR product dealing with two-dimensional formulas well.Our research group has done some work on formula recognition and published many related papers. Howerver, many jobs remains to be done and improved, such as the correcting rate of symbol recogintion and the improvement of generalization performance of the recoginzer. To this end, this thesis presents an OCR system based on neural network ensembles. In addition, we presents a classifier based on nerual network ensembles parallel algorithm in data mining field. The content of the thesis is as follows.Chapter 1 reviews the history and basic knowledge of neural network, neural network ensembles and parallel computing.Chapter 2 presents the recognizer based on neural network ensembles. Out experiments show that the OCR system has better generalization performance and higher correct rate.Chapter 3 presents a parallel classifier based on neural network ensembles in data mining.At the end of this thesis, the remaining problems in our system are analyzed. Further research and possible improvement on BP neural network parallel algorithm are also discussed.