Document Image Analysis

Document Image Analysis is our first exercise in Advance Image Processing. In this exercise, I will use OpenCV and Python.

The shaded area in the form below should be classified.

All the forms were pre-processed with given center of the circles. The circles should be classified in three different categories: shaded, not shaded, and crossed-out. 

First, the given center of the circles should be plotted from csv to the python code.  

At first, I had a hard time dealing with the syntax of python so I spent quite a few hours studying and trying out sample codes so that I can use the basic operations and functions for this exercise. I used a normal file upload then I loop the file line by line. Finally, I boxed all the circles in green lines using OpenCV's rectangle function.

After the initial boxing, the mean intensity of the boxed circles was gathered. Lower intensity means darker. The darker circles are detected blue (I assumed that the student crossed-out the circle so it is darker than the filled circle). A threshold between the black and the unshaded circle are colored by red. 

This is a sample resulting image.

I found minor errors such as filled circles with minimum intensity and those misaligned images. However, I believe that my exercise has a higher accuracy of attaining the correct answer.