Prototype inverted platen book scanner
This is the result of my first attempt for a book scanner. I scanned lots of books before, mainly
vintage computer related,
on a flatbed scanner, but this is the first attempt in scanner construction.
The main construction is 12mm MDF and acrylic glass. The acrylic glass is 320x255x3 mm.
MDF is easy to process, but not very robust. Screwing from the edge side will never make a good
The base appears to be 327x323mm, the height 270mm. Most of which results from the platen size
of 2 x A4 (297x211mm). The hight could be at least 35mm less, it is just the result of my first
guesstimate. It does affect centering of smaller scan sizes.
|With a book
Operation is very similar to a flatbed scanner, but with an angle.
|The complete scanner-camera mounting construction
The complete camera mounting construction consists of:
- an aluminium L-profile, 80x30x30x4mm, screwed onto the MDF plate, the triangular space between the profile and the MDF
is filled with a semi-flexible 3D-printed part for stability,
- an aluminium tube 50x30x20x3mm, attached to the L-profile creates some more distance (20mm) to the platen.
It is stubbed with a 3D-printed plug,
- a clamp fixed a strip (100x30x3mm) to the camera mount to the tube. The clamp is a strip of 50x30x3mm. The
padding is 3D printed.
- at the other side of the strip is another clamp and the camera mount aluminium profile, 50x30x30x3mm. The camera
mounting hole is places such that the lens center is at the middle of the mounting construction. Which itself is centered
to the platen.
The parts are mainly attached to each other with M3 flathead screws and threaded holes. Lots of work, but it results in
a clean and simple construction. The 100x30x3mm strip length makes it possible to move the camera between the centers of
an A5 and an A4 scan.
|The camera mount in parts
The camera mount screw is the standard 3/8" Withworth thread, about 0.35" long. The aluminium profile and padding
are about 3.5mm thick. The other screws are M3 flathead 10mm.
The semi-flexible filament is TPE. This is similar to rubber in flexibility.
Evaluation of the prototype
During construction and some usage a number of problems were identified:
*) one could replace the aluminium tube 50x30x20x3mm (see above) by a tube 50x30x200x3mm, but the result would be heavy, wieldy and not very robust.
- MDF plate is unsuitable for serious construction. Screwing from the edge side leads to splitting of the material,
- The distance between camera and platen is only usable for A5 sized scans, and results in some fish-eye distortion.
Some rough measurements and estimates require about 20cm more distance between camera and platen*,
- The distortion and the lines of text not being exactly horizontal also have a negative effect on the OCR quality. A rising line can even result in
(part of) words in reverse order,
- Attaching the acrilic glass with drilled and tapered holes is a lot of work, leads to the MDF edge problems and makes it harder to replace the glass. A clamping construction with similar holes, optionally 3D-printed, would be better and look cool too.
- All surfaces should be black to reduce reflections. Most or all will filter out when converting to black and white, but it still reduces the intermediate image quality,
- The current camera mount obscures the display and due to the complete setup only viewable at very awkward angles.
Mainly an issue during setup/configuration/testing, but that is my current phase.
The remainder of the setup consists of:
- Canon PowerShot A2500 camera's with CHDK firmware,
- Image processing with ImageMagick command line tools,
- OCR and PDF construction with Tesseract. Both
this and the previous are available for any mainstream Linux distribution,
The current setup doesn't have an option to redact the OCR text and reinsert it into the PDF. This would be
a very useful option
More useful links: