Design Of Automatic Identification And Score Analysis System For Examination Paper Scoring Based On OpenCV

As one of the important means of education examination,the score of written examination paper accounts for a large proportion.For a long time,there have been two ways to score test papers and manage students' scores: one is to make specific examination paper and then scan them,and the other is to deal with them manually.However,the former requires customized examination paper and special pens to write.What's more,the scanning tools cost too much to be widely used in all examination papers,while the latter use manual methods to conduct statistical scores and analysis,which is not only a waste of time,inefficient and prone to errors.In view of the shortcomings of the above marking methods,this paper realizes the design of the automatic recognition and analysis system of test paper scoring based on OpenCV.The system focuses on the recognition of handwritten Numbers and letters.As each person writes different characters and has less characteristic information of Numbers and letters,there is no context relationship between characters.A wrong character will have an important impact on the final result.Therefore,it is necessary to maintain a high recognition rate.Moreover,number and letter recognition is faced with a large number of reports and documents.If the speed is too slow,it is obviously difficult to meet the actual needs,so there are also high requirements in the recognition rate.In this thesis,after a lot of analysis and comparison of existing recognition methods,support vector machine algorithm is finally selected to identify handwritten numbers and letters,and verify the accuracy of recognition results.Based on high shoot instrument examination papers of image preprocessing,including the image threshold operation,Hough transform method is used to find interested in tilt correction,according to the boundary detection area,using projection method to extract handwritten character,character image noise reduction and morphology processing operations,such as the resulting for recognition of handwritten characters.Graphic segmentation is also one of the key points to realize the system.The paper will be graded with multi-digit strings,and the non-concatenated strings are divided by the vertical projection method.For the concatenated strings,a method based on contour features is proposed.First,the character image is refined to obtain its skeleton,and then the state machine is used to find the abrupt point of the waveform chain,that is the concatenated point between characters,so as to realize the segmentation of the concatenated strings.The system use support vector machine algorithm to recognize characters.In combination with the actual situation in marking,the number classifier and letter classifier are respectively trained.The parameters of support vector machine algorithm are selected through cross-validation method,and the feature of the direction gradient histogram of characters is extracted to classify the characters through the feature,so as to realize fast and accurate recognition of the digits and letters written by the hand.Finally,this paper gives the software and hardware environment of the system,completes the main program and interface design of the system,introduces its functional framework and operation process,and analyzes the distribution,difficulty,distinction and reliability of the final score.After several tests,the system has achieved an accuracy rate of over 96% in character recognition and a reading speed of 4 copies per minute,meeting the requirements of practical application.