| Author |
 Replies: 10 / Views: 4,948 |
|
|
New Member
United States
3 Posts |
There was some discussion about this in the past, but the person had mentioned something that I thought may have been too ambitious and unnecessarily so. They wanted an optical sorter that could read dates and mint marks...more detail than would really be necessary to get something like that off the ground.
I've been discussing with Andrew Redlon the idea that he might be able to incorporate optical sorting into his Ryedale Penny Sorter for the purpose of finding wheatback cents (1909-1958).
Even he had come up with some thinking on it that I felt was too ambitious (regarding how fast he thought it had to be).
The basic idea would be to use off the shelf components to build an adapter that could be put in place of the alloy detector which would consist of a plastic slot the pennies pass down with a hole all the way through and cameras pointing on each side of the hole. With simple OCR software, it should be able to stop the penny briefly, the camera would take a picture and send to your computer, the OCR would either verify that the words ONE CENT can or cannot be read from one of the sides of the pennies, and then route accordingly.
In my opinion, if this could be done at the rate of 10 per minute, it would be worthwhile, considering that you would first run all the cents through his excellent alloy sorter and remove the 80% or so of pennies these days which are post-1982, leaving copper cents only.
That would leave you with an amount that is far less, which you would then optically sort.
While 10 per minute does not sound like much, it does get you to 1,000 in a few hours, and in my opinion it's a great place to start to get something like that working. Speed, detection of greater detail, and diversifying into other coin types can be worked on later as the software improves. (Can anyone imagine optically sorting coins by grade?)
This would make his sorter dependent on a computer for that part of its fuction, but so what? There's no reason to make something that is standalone these days when you can have your computer doing part of the work.
Thoughts?
|
|
|
|
Valued Member
United States
374 Posts |
This sort of thing is done all the time with computer vision systems for quality control / inspection. Typically the machinery is operated with a PC, PC/104, or if using LabVIEW like many such machines do, an National Instruments PXI system. I have tackled computer vision problems like this and the big difficulty with coins is the variety of conditions which are acceptable. Quite honestly, if you're looking at a rate of 10 per minute, human inspection would be easier. I can tell you that you could achieve much higher throughput rates than 10 per minute.
That being said... I might just happen to be working on a similar project...
|
|
Pillar of the Community
United States
4809 Posts |
As mgillette has noted, automated optical inspection is already in heavy use. I would think that using it wouldn't be a large undertaking; initial investment might be tough. In my experience, shadows and contrast are the largest hurdles. Given several pennies can be in a single field of inspection, automation would have the time advantage. This assumes, however, that you could overcome the other hurdles. Grading though would be a much more complicated endeavor though.
|
|
Moderator
 United States
23522 Posts |
PCGS is already doing this with Secure Plus coins, and has owned the relevant patents for two decades. It's a great idea, but somebody's already been there and you'll get hauled into Patent Court for infringing if you do it.
|
|
Valued Member
United States
467 Posts |
A banner vision system would run you 500-1000 or so. Easy to set up, program to a diverter to kick off.
|
|
Moderator
United States
23522 Posts |
I work for a recycling firm whose mission is to create clean, recyclable glass from the Single-Stream Recycling system. Our separating equipment is optical, and easily capable of 100,000 decisions per minute. Rigging something capable of keeping up with a Ryedale Sorter is trivial. Possibly stepping on the patent prerogatives of a known-litigious company? Not so trivial.
|
|
New Member
 United States
3 Posts |
SSuperdave,
You're assuming that PCGS has the exclusive patent on optical coin sorting...they have a patent on their proprietary design, not on the entire concept in which there is room for many designs that don't infringe each other.
Would be like Microsoft having the exclusive patent on operating systems and no one could create another one.
There are certainly other ways to optically sort than how PCGS gets it done which are not a patent violation. The specific software and hardware design they use are likely patented...but no one has to use their specific software and hardware to get that same job done with something similar.
|
|
New Member
United States
3 Posts |
mgillette, I don't feel you would have to go as far as using very high tech machine vision equipment that costs $1,000s to do a basic type of optical sorting involving separating wheat cents from memorials. I honestly think that cheap, off-the-shelf cameras and OCR software could get the job done. I agree with you that there could be issues with lighting...but I think if you illuminate the target with LCDs , you could probably get enough sensitivity from an off the shelf camera to pick up enough points that it could distinguish a Wheat cent from the rest. As far as the speed issue...yes 10 per minute is slow. The goal is to get it working in the first place, which is impressive enough. After that you can double the speed and double it again. I disagree that you may as well sort them yourself. Anytime you can have a machine do what you can alrady do yourself, you free yourself to do something else. So, saying 10 per minute is not worth the effort, I would have to strongly disagree. Besides, you are not necessarily stuck at that speed forever...that's just a starting point. I look at this as 1st generation, 2nd generation, and so on. |
|
Moderator
United States
188440 Posts |
The wise thing to do is hire proper counsel and patent whatever it is you plan to do before your try to make money from it.
First, it prevents someone from stealing your idea and then suing you (since they will surely take this advice).
Second, the process should ensure your patent does not infringe on any other existing patents.
Trust me, more eyes are seeing this thread than you think.
|
|
New Member
United States
3 Posts |
If Ryedale develops this, they're not going to necessarily do it exactly how I suggest (and they shouldn't). I just suggested to them that it doesn't seem like they have to go the route they were thinking of originally, which was relying on some multi-$1,000s machine vision technology, which is unnecessarily sophisticated for such a fairly simple task. The end result of confining the idea to having to use expensive, patented technology meant for a much more sophisticated use is that it will never get done, meaning we don't have it to use, and if it did exist we couldn't afford it anyway.
I could see investing a bit in Andy's company to help him develop this 'opensource' idea, but I truly am not in this for what I
can get out of it. All I want to end up with is a reliable optical sorter that saves me having to look at the coins myself and I really don't care how fast or slow it ends up being. (If because of its non-sophistication, it never gets faster than 20 per minute...heck, I'll still buy one in a heartbeat). If you ever have a daily situation where you need to optically sort more copper pennies than can be done in 8 hours by a slow optical sorter like I propose here, well, that in my opinion is a pleasant problem to have and maybe you can afford the PCGS solution.
It's more like I'm just excited to do whatever I can to help make it a reality and if someone else gets the credit, I don't care.
What I hope to see is that his machine incorporates a simple version of optical sorting that complements the alloy detector and is done separately and as an end process after the zinc pennies are sorted out by the detector he uses. Trying to get alloy and optical sorting done all in the same process would force the detector to run at the speed of the camera, so the way it should work is that you remove the alloy sorter from his machine and attach the optical sorter and slow the speed of the coin feeder to what that process can handle.
Having to have an optical sorter of the speed of his alloy detector (250 per minute) really isn't that necessary for that job...not now anyway.
I envision having a large bin holding at least 2,000 copper pennies feeding into his machine which you've sorted out in the initial alloy detector process and leave it overnight so it's done in the morning.
That is how it could start out...but as improvements are made it could get much faster than the 10 per minute I suggested.
I take the Elon Musk viewpoint of this kind of thing...try to do it on a shoestring and improve it from there once you know what you're doing and are able to see the results after testing.
|
|
Valued Member
United States
374 Posts |
ssybesma - Almost all of this can be done with a webcam and OpenCV (BSD licensed development library ie. totally free for anything you want to do with it). Processing - feature detection, matching, blob analysis, or whatever of the many, many established and well published methods for doing what you want is fairly trivial. There are many methods of "OCR." I've worked with quite a few experts and professors who tackle much more difficult vision problems than that.
I'd have to see PCGS' patents, but they're probably not relevant to your particular use case. Then again, in my sector of work we don't deal a whole lot with patents directly (other than submitting them to our legal folks).
|
| |
Replies: 10 / Views: 4,948 |
|
Posted 02/06/2014 08:52 am
