An Automatic Marking System Based On Handwitten Character Recognition

Examination, as a kind of effective means of teaching evaluation, isaccepted by the teachers and students in the process of modern education. Aspecial sheet marking system is designed to judge objective problems, as ismainly used in China at present. This system has strict requirements about thedesign of the answer sheet. Students are required to marking the answer sheetfollow the standard rule. Specific device is needed when marking online.To overcome the shortcomings of the sheet marking system, this paperproposed a new marking system using image processing method in original testimages that is taken by any shooting device. In this system, first obtained theinformation of standard answer and student number location by matching withbenchmark paper, then positioned the region of answer and student number, andfinally a comparison in character recognition between benchmark paper andstudents’ paper in made by using AGA-SVM. This paper discusses the design ofautomatic marking system from the following aspects:In the first chapter, the research background and development trends athome and abroad is introduced, then the overall framework, the researchdirection and difficulties are summarized, finally the application prospect of algorithm used in this paper is described.In the second chapter, this paper introduces the perspective correctionalgorithm, which is used in the image of test paper. Edge-line segment detectionis designed innovatively to extract the edge-line segment of the perspectivedistortion image, then combined with perspective transform, parameter matrix isgained. Next, create structures by using projection transformation and parametermatrix. At last, combining bilinear interpolation with it to correct image.In the third chapter, how to obtain the standard answer of benchmark testand its location information and the position information of student number, also,how to use it to gain the writing area of answer and student number in test paperare introduced. Then the author use character segmentation algorithm on thebasis of concave detection of contour difference to partitioning multiple choicequiz and with redacted options on the answer sheet.The process of preprocessing the segmented image character and thefundamental principle of SVM and genetic algorithm are proposed in thischapter. Adaptive genetic algorithm was used to optimize the parameters ofSVM, then use the optimized SVM to processing character image.From the perspective of digital image-processing, an automatic markingsystem based on handwritten character recognition is designed in this paper. Notonly did it save the traditional answer sheet, but reduce the waste of paper andsave the time of full filling the sheet. Without specific and limited shootingdevice, one can mark online quickly by using any shooting devices. The experiment results show that this marking system can insure the accuracy ofrecognition up to99%. Compared with the current marking system, this newmarking system is characterized by easy operation and low cost.