Date of Award
Master of Engineering Science
School of Engineering
Dr Douglas Chia
Dr Alexander Rassau
Multiple-choice questions are one of the questions commonly used in assessments. It is widely used because this type of examination can be an effective and reliable way to examine the level of student’s knowledge. So far, this type of examination can either be marked by hand or with specialised answer sheets and scanning equipment. There are specialised answer sheets and scanning equipment to mark multiple-choice questions automatically. However, these are expensive, specialised and restrictive answer sheets and optical mark recognition scanners.
This research aims to design and implement a multiple-choice answer sheet and a reliable image processing-based scoring system that can score printed answer sheets and send back the scored answer sheets to students automatically. The proposed method will allow users to print the proposed answer sheet and use a normal scanner and computer to perform the scoring. Moreover, while compiling the assessment results, it will annotate the answer sheets with feedback and send back to students via email.
Firstly, the proposed system requires class list and scanned answer sheets to perform scoring. Then, the software algorithm will first locate the finder pattern points and correct the answer sheet tilt. After recognising the finder pattern points, the algorithm processes student information in an aim to match the decoded student information with the class test. After accomplishing student information recognition, based on the key answer, the algorithm scores the answer area which contains 72answer areas and five choices in each. In the last stage of the processing, the algorithm annotates the answer sheet and store the score in a spreadsheet.
There are three major research contributions. First, a new method to change multiple answers on the answer sheet. On the proposed answer sheet, students do not need to use pencil and eraser to change answers. Second, a new method to recognise student information on the answer sheet. The proposed system replaces the conventional method of encoding student information on answer sheet by introducing new method to decode student ID and utilising the count of student name characters. Lastly, a new fast method to provide a test result feedback. The proposed system provides annotations on the answer sheet and sending the answer sheet to student email.
After experimenting the system, the system results indicated that the system meets the research objectives. Speed wise, the system scoring speed is 1.25 seconds without annotation and 2.25 seconds with annotation. The software algorithm proved to be robust to detect the finder pattern points when noise exists in the answer sheet margin and pen scribbles around the finder pattern area. In addition, the algorithm is able to correct scanning tilt up to 5-degree rotation. Furthermore, the algorithm is capable of recognising different shading styles as long the shading area covers at least40% of the answer box. A case study was conducted in a real test situation to retrieve more results about the system. The outcomes of the case study are that the success rate of finder pattern recognition was 100%, the success rate of marked answer recognition was 90.2%
Alomran, M. (2018). Automated Optical Mark Recognition Scoring System for Multiple-choice Questions. https://ro.ecu.edu.au/theses/2137
Available for download on Thursday, December 14, 2023