Font Size: a A A

Research And Implementation Of Printed Russian Character Recognition System Based On The Neural Network

Posted on:2008-12-10Degree:MasterType:Thesis
Country:ChinaCandidate:R HouFull Text:PDF
GTID:2178360215459346Subject:Computer software and theory
Abstract/Summary:PDF Full Text Request
With the rapid development of Chinese economy, the recognition technology for the printed character is applied widely. Moreover, a lot of relevant departments and enterprises bring forward many new demands for it. Particularly, in the recent years, the number of import and export between China and Russia increases steadily, which increases workload of handling documents. Thus, the application of the character recognition technology will doubtlessly promote the operation of the Sino-Russia border trade. However, in China, there are a few institutions researching the Russian character recognition. Based on the above reasons, the research in this thesis aims at the printed Russian character recognition.This thesis discusses each preprocessing ways in the printed Russian character recognition system, studies the method of feature extraction from Russian characters, and designs the classifier of the Russian character recognition system. Pictures to be recognized from a scanner are acquired. The special areas in the pictures, which are binarized by fixed threshold based on the principle of the combination between global threshold and local threshold, is extracted.The threshold of line segmentation through means method is dynamically chosen, and the line segmentation for the Russian character recognition areas by the level projection and threshold method is completed. The vertical projection for segmenting Russian words on the basis of great difference between Russian word space and character distance is adopted. On the basis of successful segmentation of Russian words, to further segment Russian characters, the vertical projection and threshold method is used. Moreover, Russian character features by pixel is extracted.This thesis adopts the BP neural network as the classifier, and designs the BP neural network structure in the Russian character recognition system. Then the BP neural network is trained .The BP neural network completes Russian character recognition. The printed Russian character recognition system is realized by VC++.
Keywords/Search Tags:Printed Character Recognition, Russian Character Recognition, Character Segmentation, BP Neural Network
PDF Full Text Request
Related items