ACCURACY ENHANCEMENT OF OMR FOR EXAM MARKING WITHOUT PRE-SET TEMPLATE USING IMAGE PROCESSING

Optical Mark Recognition (OMR) is used to automate answer matching
especially in the education sector. OMR marking machine is costly and limited to
specific OMR paper design, thus launching researches using image processing to find
less costly solutions. However, studies so far have achieved relatively low accuracy
and poor consistency unless a fixed OMR form design is used. Accuracy drops with
more OMR questions. Therefore, this study investigate means to improve OMR
marking accuracy using enhanced algorithm designed for OMR marking. The results
were compared against manual marking as the control and existing image processing
algorithms. The metrics used are F1 score and percentage error for accuracy of
detected answer options and marking fault respectively. The result is encouraging with
consistent full accuracy for up to 90 questions as compared to previous works.