A Information Card Recognition System Based On Support Vector Machine (SVM)

Recently, with the development of education, thetechnology of reading answer papers automatically based on OMRtechnology has been widely used in National MatriculationExamination, Adults' National Matriculation Examination,College English Test Band Four/Six , Computer levelexamination , etc. But there are many limitations intraditional method, such as equipments are not be all-purpose,the format of information is fix-up, the cost is high, theprocess of correct faults is complex, images can not beconserved and so on. So developing a new algorithm to resolvethese problems has become a research hotspot. Based on thesecharacteristics and the limitation of current method, thisthesis has done mainly the work of three following respects.Firstly, this thesis deeply studies the essentialissues, related problems about documents images recognitionbased on optics sign and the process of pretreatment forinformation card. Based on Hough transform algorithm, aimproved algorithm which is used to measure the slope angleof image for information card has been put forward.Secondly, we study the Statistical Learning Theory (SLT) and Support Vector Machine (SVM) Theory seriously, and discuss classifying algorithm and kernel function. This thesis expatiate the research and application status of Support Vector Machine , and propose a new algorithm based on SVM for information card recognition algorithm . It can recognize different images exactly which are obtained at different situation such as different handwriting, different material of paper, different shape when it moves.Finally, according to the new algorithm this thesis design and develop a system which can reading all kinds of test papers intelligently. This system not only can read answer paper intelligently but also can analyze the result and form the record report. So the teacher can teach students in accordance of their aptitude after finding the weakness though the report.