Hey all,
Has anyone ever altered an R package for image analysis to do optical mark
recognition? I'm trying to find a way to semi-automate data entry of
several thousand paper health surveys that are predominantly composed of
check boxes. All the boxes are uniform size and shape, so it seems as if
it should be possible to alter a package to recognize the boxes and output
a 0 or 1 to correspond to whether the box is empty or not. From there, I
could write if/then statements to convert the output into the relevant
answers to the questions in the survey, and export it for analysis.
I've been playing around with imageHTS and EBImage, but haven't been
able
to alter the configuration files that come with the packages for use in
identifying cells, electrophorisis screens, etc. The surveys all have a
line down the middle and an ID number on each page, which should serve as
anchors for the program as it identifies the boxes.
My question is whether or not anyone has ever tried anything like this, or
if you even think it can be done. Having spent a week on this, I'm
starting to doubt my initial assumption that it should be an easy
alteration to make.
Thanks for any input you might have,
Kirsten Simmons
--
Kirsten Simmons, MPH
Polymath interested in productivity, how ideas spread, gardening,
marketing, entrepreneurship and models of social networks and disease
transmission
[[alternative HTML version deleted]]
