Simple, easy to build, ~600pg/hr, very portable scanner

[Mohib]

Ok, as mentioned over in David Landlin's thread here:
http://www.diybookscanner.org/forum/vie ... 110#p17661

I had ordered one of this 10W LED lights from Amazon:
http://www.amazon.com/Waterpoof-Outdoor ... roduct_top

It just arrived and of course I had to try it out, so don't take these preliminary results as anything more than a quick assessment evenness of the light distribution and a first attempt at colour. These were taken at night so there was no daylight.

First impressions, the light is very well built, much more compact than I expected, impossibly bright to look at directly (as you can get the idea from the reflections in the pictures below) and is a good general light but is probably not bright enough for most people to get the kind of results they'd want from good book scanning. My impressions are 2x20w will probably satisfy most people. That being said I'm happy with this lamp because keep in mind that my overall objectives are quite different from most who will be looking for more perfect results and have more elaborate scanners and cameras. My objectives for this lamp were:

a) Get more even lighting across the page than I was getting with ambient light or a clip lamp to the side so ScanTailor generates more evenly weighted text (otherwise it generates somewhat heavier letters in the less well lit parts of a page) as my primary objective is high quality OCR not archival or other purposes.

b) Eliminate the reflection of the camera, camera arm and other items in the image (as can be seen in the images below taken from before).

c) Given the limits of my rather simple camera (an Olympus VG-130 14MP http://www.olympusamerica.com/cpg_secti ... sp?id=1533 ) try and get a reasonable colour rendition if there is a white balance setting that understands this light.

d) Add a lamp to the scanner with virtually no handiwork.

Now over all, I really don't need the light because ScanTailor does a phenomenal job, but I would like to have a reasonably good original should I need it for some future purpose.

WRT my objectives:

a) A more even light is produced if it is tilted up, away from the platen and towards the camera rather than (as in the picture) pointed right at the platen. I'm happy very with the results compared to what I was getting before and the lighting is much more uniform than it used to be and probably as good as I'm going to get it with 1 $10 lamp and my other design constraints. My objective here has been met.

b) Clearly there is a problem with larger originals as the light is reflected in the image (and creates a secondary purple hot spot). But if the platen is moved 2.5" away from the main post then the problem is solved, so I will just cut a new pipe for the camera arm that's a little longer. Otherwise all other reflections are gone and my objective here has been met.

c) Colour is not half bad on the auto white balance setting of my camera. The brown paper wrapped over the book cover (you can just see the edge of it to the right of the page in the 3rd picture showing page 81 of the book) is very close to the actual colour. I'll try again in the morning, when there's daylight, and see if I can do better with the mixed light. Otherwise, I'm happy with what I'm getting compared to what I used to get (which was basically worthless colour). My objective here has been met.

d) Well, as shown, there was no handiwork but a simple u-bolt over the top of the T-Joint going into the light bracket is all that it will take to mount this light and keep the design otherwise unchanged and avoid extra paraphernalia. My objective here has been met.

The pictures below are:

1) The temporary set up I've got showing where the light is placed (using the temporary bungee cord mounting). The light is about 14" away from the surface of the platen.

2) Result I used to get with a small original (standard book) before I got the lamp in daylight (but wrong white balance setting). Note the reflections

3) Result at night with a small original (standard book) using the lamp with the camera set to auto white balance setting. Colour is actually quite accurate, no reflections and fairly even lighting.

4) Result I used to get with a larger original (letter size) before I got the lamp in daylight. Note the reflections.

5) Result at night with a larger original (letter size) using the lamp with the camera set to auto white balance setting. Colour is actually quite accurate, lighting could be a little more even, and platen has to be moved 2.5" to remove light reflection.

6) Result at night with a larger original (letter size) with colour using the lamp with the camera set to auto white balance setting. Colour is actually quite accurate, lighting is not bad, and platen has to be moved 2.5" to remove light reflection.

7-10) Result at night with a larger original (letter size) with colour using the lamp with the camera set to various white balance settings: Cloudy, Fluorescent, Incandescent, Sun.

Well hope this helps people get an idea of what one 10W LED bulb does.

Overall I'm very happy with the performance. It's let me meet my objectives better keeping in mind my design constraints.

Light location held in place temporarily with bungee cord

Small original - Without lamp in daylight but wrong white balance setting (original result before lamp)

Small original - Auto white balance

Letter size original - Without lamp in daylight (original result before lamp)

Letter size original - Auto white balance

Letter size original with colour - Auto white balance

Letter size original - Cloud white balance

Letter size original - Fluorescent white balance

Letter size original - Incandescent white balance

Letter size original - Sun white balance

[Mohib]

I've made some more controlled tests to get a better feel for colour and light distribution. My tests were of a book 11x13" (the limit my scanner can handle, but the stepped zoom didn't allow framing the book well, so the field of view was either too wide or too narrow and about 10% of the book around the edge was missing).

The lamp is attached similarly as shown previously, but is now hanging loose from the camera arm rather than pulled tight up to it (there really was no need for that and this set up allows the lamp to be removed easily so I really don't think the u-bolt will be needed) so it's a couple of inches closer to the platen. See picture below.

Lamp hanging on bungee cord.

One thing I noticed was no matter what sized book I used, the lamp didn't really show up as a reflection like it was last night -- I think this is because I had the camera tilted away from the post a little (so that will introduce a bit of key-stoning) and so the lamp stayed out of the view.

I've also decided on a small change and have ordered from ebay a 6" macro focusing rail (shown below) which I'll attach between the end of the camera arm and ball joint. This will allow me to easily move the camera up and down so I can frame the material better, irrespective of the camera's zoom. It will also partially resolve the page scale problem mentioned earlier as every 20 pages I can move the camera 1mm and keep roughly the same on thick books.

6 inch macro focus rail - small.jpg (30.8 KiB) Viewed 8023 times

The remainder of the pics are tests. All were done on Auto White Balance as that seems to give the best results. The first picture which is the benchmark, i.e. no lamp. There are two pictures in each test. The first with daylight coming through cloth blinds that are closed, and the second with the blinds fully open and daylight. The latter seem to be better, with more even lighting and better colour due to the extra light. Of course smaller originals have more even lighting than larger ones, but a regular magazine is pretty good.

Overall given my constraints the light has made the images much, much better than I used to have and I'm able to get decent pictures at night as compared to using an incandescent clip lamp and it's eliminated all reflections. It also fits into the design nicely without increasing the footprint or complexity, although I don't see using it in a library, but it's compact enough to take along and try. Before good colour was an option, now reasonable colour is and it seems the best choice will be to image with the lamp in day time.

The lamp comes with a short 9" cord and right now I've just hot wired it to a short wire I had and plug that was already attached, however my plan is to simply attach the lamp to a switch and the other end of the switch to 15' of wire. That will avoid hotwiring and also put a switch in which will be handy. After about 30 mins it just gets mildly warm.

Large 11x13" book. No lamp. Colour is whiter and cleaner. Fairly even light, but note the reflections.

Large 11x13" book. Lamp. Colour is bluish. Lighting is not as even and right side of page (near the lamp) is a little washed out. No reflections.

Large 11x13" book. Lamp + full daylight. Colour not as bluish. Lighting is more even and right side of page (near the lamp) is not as washed out. No reflections.

Regular sized magazine. Lamp. Colour is a little bluish. Lighting is not bad and right side of page (near the lamp) is a little washed out. No reflections.

Regular sized magazine. Lamp + full daylight. Colour not as bluish. Lighting is more even and right side of page (near the lamp) is not as washed out. No reflections.

Small 9"x8" book. Lamp. Little blurry for some reason. Colour is fairly clean. Lighting is fairly even,but at the right side of page (near the lamp) ink on the page is reflecting the light creating the circular pattern. Increasing the length of the camera arm, and thus changing the angle of the light to the page to the camera solves this. No reflections.

Small 9"x8" book. Lamp + full daylight. Still a little blurry for some reason. Colour is fairly clean. Lighting is fairly even, and reflections of the ink at the right side of page (near the lamp) are less pronounced. No reflections.

[cday]

The LED lighting looks very promising, but areas of the page that are nominally white look light gray, have you tried increasing the exposure time? That was suggested by Daniel in another thread recently...

[Mohib]

Thanks for the suggestion. Yes, I tried increasing the exposure time but then right side of the page gets more washed out and the text becomes very thin. In fact, on scans where I'm not concerned about colour and will be using ScanTailor to drop out the background, I'm actually leaning towards dropping the exposure by 1/3 stop as that thickens up the text nicely for ScanTailor and OCR.

Of course my real problem is working with one light and why the lamp plus sunlight combination gives the best results as the sunlight somewhat evens out the overall illumination. Further, I'm only using a 10W bulb and, as has been noted elsewhere, 30w - 40w of illumination makes a huge improvement in image quality. However, these deficiencies are unavoidable given my objectives and design constraints and what makes my project an exercise in compromise and adaptation -- not to mention much thinking! -- depending on the purpose of the scan.

I thought about using a 20W bulb, but they are much bigger in size and I'm unsure as to whether it will wash out the side of the page closer to bulb even more or if that would remain the same relative to the other side of the page given the exposure would change accordingly due to the extra light.

But, as you say, the results look promising and much can be done to improve the results on "conventional" projects: 2 or more lights, more powerful lights, cameras with manual control over white balance, etc.

Nevertheless, for me, the light has made my scans much, much better compared to when I was using just ambient light or a small incandescent clip lamp and also removed the reflections (although there is still a slight hint of one in the sky area). And whereas, before I thought scans where colour is important would not be a practical option, they now are.

[cday]

Not sure why brighter light should lead to thinner text... Is the paper of that book quite shiny, as it looks as if it might be? Have you tried with a more typical book paper? If it is simply a thresholding effect at the camera sensor it could indicate a lack of pixels, but that doesn't look to be the case.

In general, of course, the illumination from a lamp is more even the further away it is, so a brighter lamp further away would provide better illumination. Not so practical for your novel design, though.

[TomHorsley]

I built a light system that consists of a little current regulator module and 4 Cree LEDs, each stuck (with a chunk of aluminum for a heat sink) on the end of a stalk of heavy copper wire so I can bend the wire into position. Two on each side of the book positioned to avoid direct reflection into the camera seem to work quite well. I've considered some Styrofoam balls for diffusing the light if necessary, but I'm not sure I'll need that.

[Mohib]

Ok I did a bunch of tests with 4 books -- Dakshin, Trees, 1001 Inventions, Golden Prince (which I have in separate 10MB zip files but they won't seem to upload, so I've attached a few samples) -- using the lamp, with my blinds closed (you can see what they're like in the 3rd pic in my opening post in this thread), and it's a very grey, overcast day so the primary light is going to be the LED lamp.

I did one test (Dakshin) without the lamp just as a reference.

All books have smooth matt paper -- coffee table type material except the Golden Prince which is a 1950s on rough, almost card like paper.

1001 Inventions and Trees are large books -- 10" & 11" wide respectively so the right edge is closer to the lamp. Dakshin is 8" (typical magazine width) and Golden Prince is a typical book 6" wide.

I just eye-balled the camera level so there may be some distortion (although I know the lens does introduce pincushion effects which is noticeable, especially when it's perfectly levelled and square, and particularly noticeable in on the right margin of text in the Golden Prince).

Camera was set to 5MP and auto-white balance. Colour balance is a bit on the bluish side, however the Golden Prince which is very good (the brown cover is almost spot on). Presumably with a white page and no over powering lamp in the field of view the auto-white balance could do it's thing better.

Each test set goes from -1 stop, -.66 stop, -.33 stop, 0 , +.33 stop, +.66 stop, +1 stop.

And from what I can see, CDay, you're right. Generally over exposing a bit doesn't make the text thinner (well marginally perhaps at +1 stop) on the right side, except on wide books where that side is closer to the lamp. But otherwise the scans are much better and brighter. Thx for the tip and not sure why I was thinking the text got thinner, but must have mixed up some earlier tests and got confused.

However, is it my imagination or is the lamp messing up the autofocus system? Or is it what looks like a halo around the letters that's making it look like less sharp? Because to me the Dakshin pics with the lamp do not seem as well focused or crisp as the ones without.

I've thought about mounting a second lamp (a bike headlight) on the handle next to the break lever, but they don't have a wide dispersion and are more like spots so I'm not going to bother.

I'm going to experiment with raising the light back upto the top and lower down just to see what difference it makes.

Tom, I tried various diffusers too but they didn't really seem to do much good or bad or anything for that matter.

Dakshin (no lamp) -0 stop-small.png

Dakshin (no lamp) +.66 stop-small.png

Dakshin (no lamp) +1 stop-small.png

Dakshin -0 stop-small.png

Dakshin +.33 stop-small.png

Dakshin +.66 stop-small.png

Dakshin +1 stop-small.png

Trees -0 stop-small.png

Trees +.33 stop-small.png

Trees +.66 stop-small.png

Trees +.1 stop-small.png

1001 Inventions - 0 stop-small.png

1001 Inventions +.33 stop-small.png

1001 Inventions +.66 stop-small.png

1001 Inventions +1 stop-small.png

Golden Prince -0 stop-small.png

Golden Prince +.33 stop-small.png

Golden Prince +.66 stop-small.png

Golden Prince +1 stop-small.png

[Mohib]

Well after much foot-dragging, I've finally completed the construction instructions and schematics to build the scanner.

I've called it the 7 by 7 scanner because when it's all said and done, it just needs 7 cuts in plastic pipe and 7 pairs of holes to be drilled to make it!

It's a first draft so please excuse the inevitable typos. If you spot any I'd be grateful if you could let me know along with any feedback on parts that are not clear.

Apologies for the lack of pictures in the instructions themselves, but I didn't take any while I made it and really didn't want to make another. But, really, it's rather straight forward, although I go into every nitty-gritty detail because some who have asked me about building it clearly are not into handiman work.

I've got a v1.1 and v1.2 of the scanner in the works and will post the design changes once I'm happy they work.

BTW I did finally get a circular bubble spirit level (from e-bay, from China for $3 with free shipping as surprisingly these things aren't available in local hardware stores) and it really does make a night and day difference in levelling the camera. Of course the table may not be level so you have to check the table and then match the camera to that rather than really making it level.

The documentation is in 2 parts because I can't seem to upload it as one file here. Part 1 are the instructions and bits and pieces needed to make the scanner. Part 2 has the schematics.

2014.05.13 - 7 by 7 Bookscanner (V1) Construction Instructions & Schematics - Part 1.pdf

(6.84 MiB) Downloaded 524 times

2014.05.13 - 7 by 7 Bookscanner (V1) Construction Instructions & Schematics - Part 2.pdf

(2.04 MiB) Downloaded 378 times

[Mohib]

Well right after I posted the documentation, my new parts for some of the changes I was considering came in and since they worked well, I thought I'd post the pics here and will update the docs in due course once my final mod is done which is a semi-custom part that will simplify the camera mounting hardware as well as making it more flexible so almost any kind of small camera will work because right now the trigger and camera mount must be "custom" made for each camera.

My 3 changes are:

1) The bungee cord is gone! In place of it and the eye-bolt, I run a length of 1-4"-20 threaded rod through the camera arm that attaches to the ball joint/macro focuser at one end, and at the other end, the rod goes through a large washer that is the same diameter as 1 1/2" pipe (which I happened to find in HomeDepot's nuts and bolts section) so it drops into the back of the T-Joint and rests in and against it just like a piece of pipe. Another, smaller washer takes down the hole size and then wing nut tightens it all up pressing the large washer into/against the T-joint. You can just see the top of the threaded rod sticking out from the T-joint in the last pic. I'll cut it shorter (so it's hidden inside the T-joint) once I've finalised the camera arm length. The rest of the camera-arm hardware is as per the docs. This set-up works a lot better than the bungee cord, as the whole top end is now solid and it's still just as fast to take apart. It also looks a lot better! The down side is because the top end is now clamped up tight, the camera end is a little more springy than it used to be as the tinly slop in the camera arm parts and bungee cord would dampen out vibrations like a shock absorber. But it's only marginally more springy and it's just as stable while scanning.

2) I got the new Osrso ball head ($8 incl shipping from ebay: http://www.ebay.com/itm/Osrso-FRT-02-Bl ... 2a2fe21427 ) mentioned in the docs. It's much better than the Pedco. Very well made, almost as solid, but most importantly doesn't shift when tightened. I've removed the camera base from it so it threads further into the hex-coupling nut and gives a more solid connection.

3) I've added the macro focusing rail ($11 incl shipping from ebay: http://www.ebay.com/itm/2-way-Macro-Sho ... 0912760547 ). This lets me fine tune the camera height so I can fill the frame nicely as the zoom on these small cameras are often move in very wide steps. I can only take it about half the travel down because then the platen handle hits it while tilting. Making the platen handle 1" shorter I don't think would affect it's function or ergonomics much and allow a little more travel. A taller main support post is an option, but a little more vibration will result. In any event, the slot in the focus rail allows me to make course adjustments with the camera and allows an additional 4" of travel.

Total length of the Osrso and macro focusing rail is about the same as the Pedco ball head, so the camera is in basically the same spot horizontally. I will be trying out a 1"-1 .5" longer camera arm just to help eliminate the light reflection, but I've basically decided to let the light hang off the bungee cord as shown earlier, and at a lower level if needed rather than moving the camera out much further just to avoid introducing more springiness. If the light is lower, it doesn't reflect, although the spread across the page is a little less even, but that's the compromise. Overall it does it's job well as explained in the earlier posts (except for glossy/magazine type pages where these circular rings -- see the cook book picture posted earlier -- appear on photos near the hot spot that I can't figure out where they're coming from as they're not on the platen or the light).

The next option to stiffening up the whole assembly is to go larger diameter pipes, as pipe stiffness is dependent on diameter more than thickness, although a schedule 80 ABS pipe at the current diameter would also be a worth while test first.

And, finally, you can see the circular bubble level on the camera. $3 incl shipping (!) from ebay: http://www.ebay.com/itm/Round-Circular- ... 43c7eb90fb . Works like a charm, especially with the Osrso ball head.

7x7 v1.1 - with macro focusing rail & no bungee cord.JPG (400.34 KiB) Viewed 7878 times

7x7 v1.1 - macro focusing rail & circular bubble level.JPG (420.2 KiB) Viewed 7878 times

7x7 v1.1 - with macro focusing rail & no bungee cord but thread rod through camera arm.JPG (351.16 KiB) Viewed 7878 times

[Mohib]

I've updated the scanner to use a smartphone camera and bluetooth trigger. This simplified construction tremendously and increased throughput to 900-1100 pages per hour.

Details in this thread:

viewtopic.php?f=14&t=3401&sid=cc3b60a08 ... 18b0037ca7