Research On Business Card Recognition Method Based On OCR Technology

Business card plays an irreplaceable role in the daily life and work, it has alreadybecome an important status information carrier.Because ordinary business cardcontains the name, address, telephone number and internet site information, So mostof the cards for business cards which in Chinese and English mixed characters. Atpresent, speed is slow and accuracy is low are the questions of Chinese and Englishmixed business card recognition. This article attempts to solve these problems bystudying some new technologies on OCR.The paper in allusion to the problem of binarization methods in charactersrecognition, such as global threshold method accuracy is low, local thresholdbinarization method produces artifacts and run slower, and combination of globalthreshold and local threshold method is less effective when dealing with complexlayouts cards, this paper researches an improvement binarization method on the basisof combination of global threshold and local threshold, the method uses globalthreshold method to calculate the optimal global threshold, the choice of globalthreshold method binarization the pixel which far away from threshold; the choice oflocal threshold method binarization the pixel which closer from threshold, it makesthe business card image binarization results more clearly. Secondly, this paperanalysis the card document layout analysis using expansion algorithm which based onmathematical morphology, to complete division of the image layout blocks. Use theprojection to determine the attribute of layout blocks, extracting the text block, it cananalyze complex business card document layout quickly and accurately. Then, thereare some problems such as incomplete segmentation and lower recognition rate forChinese and English mixed character recognition, the article has researched animproved segmentation algorithm, this paper provided an unit merging algorithm offeedback recognition, characters of left-right structure incorrectly break up as Chinesecomponents and merging them. Experimental results show that this method is superiorto conventional components merge algorithm. After components merging, detectionthe character adhesion and re-segmentation, character recognition rate is increased. Business card information classification on the basis of heuristic rules-basedinformation classification algorithm, the paper use layout information in images toimprove automated categorization for text information in business cards, the accuracyof the classification of text messages is greatly improved.This paper uses the proposed method for testing business card recognition andcomparative with the results of the existing test methods. Reach conclusion theresearch on business card recognition method based on OCR technology have highrecognition accuracy, low complexity and fast, it can be applied to a variety ofbusiness card layout.