Restoring Old Paper Tape

[Charles Anthony]

On Thu, Jan 21, 2016 at 6:21 PM, Jason T <silent700 at gmail.com> wrote:

> On Thu, Jan 21, 2016 at 8:04 PM, Charles Anthony
> <charles.unix.pro at gmail.com> wrote:
> > For part 2, personally, I would take movies of the paper tape moving and
> > doing image analysis to recover that data; this occurs to me because I've
> > done a fair bit of image recognition software, so this solution may not
> be
> > feasible for all. If you sent me a sample movie, I would make a stab at
> > writing some data recovery software.
>
> I have heard of those approach and was thinking it may be a solution
> in cases where the tape is too fragile (and that's pretty likely
> here.)  It would be well beyond my abilities but might make an
> interesting project for you or anyone else with the skills.
>


The general approach would be to have the tape backlit (on a piece of
glass, with a light source and and diffuser underneath ) and  guide block
that the tape slides against so the holes move left-to-right but not up and
down. The camera is set up so that the tape fills the image as much as is
feasible. You start the camera, and slide the tape. Constant speed is not
important, but avoid backing up.

Grab a frame from the movie. Figure out the approximate pixel coordinates
of the data and pin feed holes in the axis moving across the tape (eg, the
1 bit is about 24 pixels from the top of the image, the 2 bit is about 47
pixels from the top, etc).

Process the movie a frame at a time. Grab a column of pixels from the
center of the image from top to bottom. Look at the pixels around where the
pin feed is, decide if they are light or dark. If light, the a character is
centered in the column. If not, move to the next frame. look at the pixels
around where each data bit is, and decide if the are light (punched) or
dark (unpunched). Write out that data. Skip frames until the pin feed
pixels go dark, and then skip frames until it goes light again; that will
be the next character. Repeat.

The pin feed holes greatly simplify the process. This process is quite
analogous to reading multi-track magnetic media with a timing track.

Test on a known tape. Debug. Run over damaged tapes; data recovered.

-- Charles