Reserch On Image Processing Technology In Intelligent Scoring System

Compared to the traditional manual marking,the results of online marking are more accurate,more efficient,marking process more secure.Online marking system was introduced in the 1990 s,after decades of development,it has been widely used in large-scale examination.However,most of the small and medium-sized examination still using manual marking method,in large-scale examinations,also need to use the optical marking reading,which costs much more,maintenance more difficult,and depending on the card separation model.In the case of single card mode,you must use the complex template carrying all the information to analyze the papers,take up a lot of resources,and extend the marking period.In addition,the segment of the text paper image is mostly fixed area division,not according to the actual answer area of the candidates for intelligent segmentation.Based on the above issues,this paper choose image processing technology in intelligent exam scoring system as the subject,emphasizing on image preprocessing and identifying the information of test paper,mainly studies the following related content:(1)Two kinds of tilt correction algorithm were realized.One kind is for the test paper have a synchronization head sheet,this article proposes a new algorithm for this mode.The second kind is for small and medium-sized exam which have no synchronization in the answer sheet,using the method of image texture recognition to get the tilt angle.Experiments showed that the two algorithms can meet the demand of serious form of the sheet and improve the efficiency.(2)The location is roughly determined by using lite XML template.Lite XML template only contains the point coordinates the whole choice questions area and subjective questions area,not the size of the region in detail.In addition,eliminating the step of the DOM to parse the XML file,greatly reduce the burden on the system.(3)A simplified information and identification method was achieved.Considering that in traditional exams,examinees need to manually fill personal information,and the problem of character recognition error,this article uses a one-dimensional bar code record examinees information.Experimental results show that the method makes the process of examinees information recognition more easier,improves the identification accuracy efficiently.(4)Marking of objective questions was achieved.The objective question is twice divided,than compared with the OTSU threshold to get as a reference,the chosen option could be found.This method can solve the problem of objective questions identification in many marking system.(5)A new subjective questions recognition algorithm is proposed.In the subjective scoring,a horizontal projection method combining the vertical erosion operation is proposed,and a more intelligent algorithm based on the actual answered area is achieved.All of these algorithms formed a relatively complete image processing theory system in intelligent scoring.It was concluded from the experimental and data results that the algorithm is effective in solving the problems left in the system of scoring,and the performance of the algorithm is improved.The design of the test papers is also more flexible,providing the examinees and teachers more convenient conditions.