Understanding Of The Test-paper Based For The OCR

Traditional ways of examination papers costs lots of man-power, materialresources and long time. It’s management is not easy. With the development ofscience and technology, people’s time ideas become stronger, it required the cycle ofmarking paper more shorter, so that the way of network marking are being appliedrapidly.Recently image processing technology rapidly develops and is used in the OnlineMarking System. Online Marking System is a huge system which is mainly includesImage acquisition, Image segmentation, Questions distribution, Online scoring,Results of the recovery and Analysis module.Because image segmentation module is an important step in the network scoringsystem and there will be a large number of papers need to be segement.This paperproposes the method which is understanding of test-paper based for the OCR. Thismethod is combination of commonly layout analysis algorithm with the OCRcharacter recognition technology, which is used in segmentation and understanding ofimage layout of examination papers. The method is mainly focused on the analysis ofthe paper layout. Firstly, it uses commonly layout analysis to segment thepaper.Secondly, it uses OCR technology to verify the result. Thirdly, it manages theresult of layout analysis.For several key parts of layout understanding,The research work are asfollows:In the part of layout analysis,the paper uses iterative projection to completeeach row and column segmentation and then carry out the other necessarypretreatment. A priori knowledge is combined with the connecting region method isused to accurately determine the location and ensure that the character region isintegrated by corresponding connectivity region merging. This paper uses the Fourierboundary descriptors as the main feature for character recognition, and combines withthe other characters to identify in the OCR module.In the part of layout analysis, this paper combines with the projection of layoutanalysis and OCR recognition results to layout analysis and classifies the questions,positions and identifies Objective questions.The results of experiments show that thealgorithm proposed by this paper is very effective. Tests found that the iterative projection method in the early layout segmentation can achieve the anticipated effects.The recognition algorithm based on the Fourier boundary descriptor character for theidentification of the question number also meets the requirements. So from the overallsystem which we build is comprehension meet demand.