Optical-Mark-Recognition-Android

OMR Scanner android app made using OpenCV Android SDK capable of recognising simple omr sheet bubbles and generate the result using an answer key.

Uses OpenCV contour detection and Image Processing to recognise darkened bubbles and prints the recognised answers with 95-100% accuracy.

Takes an image of the OMR sheets and gives an option to crop it to make it fit the given discription and also to remove background noise.

Saves it and then scans it using OpenCV and saves the results in a file which can then be checked with an answer key that was added prior to the scan.

Can save an answer key and also modify existing ones within the app.

Can automatically check the answers with the defined answer key and grade individual scanned sheets.

Can even scan an OMR directly from the computer screen in case of digital records with a good accuracy as well and provide similar results to a real OMR sheet.

Can recognise upto 20 question blocks and more.

Uses Otsu's Thresholding method to invert the foreground and background colors to detect filled bubble with ease and less effort.

This is an example on how it works

See the app in action.

Working of the app on 2 different OMR sheets taken stright from the computer screen.

https://youtu.be/g2s8hGLgqaw

https://youtu.be/6ZmqwAGwBTQ

Why not use circle detection?

Tuning the parameters to Hough circles on an image-to-image basis is a difficult task.

Many times, whether purposely or not, we have filled in outside the lines on your bubble sheet. Circles don’t handle deformations in their outlines very well — the circle detection would totally fail in that case.

Because of this, it is recommended using contours and contour properties to help us filter the bubbles and answers and then process them the way we want.

More information on the method and algorithm used can be found here: https://www.pyimagesearch.com/2016/10/03/bubble-sheet-multiple-choice-scanner-and-test-grader-using-omr-python-and-opencv/