Data extraction from exam answer sheets using OCR with adaptive calibration of environmental threshold parameters

Manual Data Collection from a student´s exam-sheets is always a tedious job which exacts ample amount of time and effort. This paper has suggested a novel approach for developing an automatic, adaptive, fast and reliable system capable of recognizing enrolment number and corresponding marks of student from answer sheet and storing it in the host computer. This system consist a hardware which picks out sheets one by one from a bundle and captures image of the front page of each answer script. This image is processed by proposed robust extraction and noise removal algorithm adaptive to environmental conditions. It is then passed through Optical Character Recognition (OCR) system which extracts characters using correlation. Accuracy of system depends on the sample space size of OCR system. In our experiment we have archived average 81% accuracy in various light and paper (Exam-Sheet) condition. We had trained the OCR with 50 samples of numerals set (0-9). In this way developed system will not only replace the traditional tiring way of manual writing of marks in database but in addition can calculate average marks of all students, ranges of marks for assigning different grade and provide grade for each student automatically.