Machine that sorts coins by date

Just tryin' to help you raise some working capital :thumb:

 

Log in or Sign up to hide this ad.

I can't see where the financial community would need this. It would have a very limited market to large dealers and hopeful collectors. If the price exceeds $3,000 for dealers or $500-$750 for collectors, you have reduced your sales potential considerably. I don't think the demand would continue either. Once bought by the few who would actually need it or use it, the market shrinks to a degree where sustained sales are almost impossible. Too risky and no guarantee.

 

If it spit out rolls/partial rolls for each individual date/mintmark with the capability to run thousands of coins(enough to make that kind of sorting worthwhile) I would see a use with large scale dealers. That is talking a lot of money, also that would mean a top and bottom scan of the coins as to not have to flip them in machine.
The problem is you dont want to have to run your coins thorugh such a machine 100-200 times just for a complete sort, think about the wear on uncirculated coins the bin and sorting could cause to a valuable coin. If you are talking about for personal use you could use low res scans, size detection for denomination and if the right size just search for asked for date pattern, would only be checking 3 things, the last 3 of date and mintmark. I dont see checking that a penny is a penny, just checking that something resembling the patterns is in the right location. the key is to keep it simple in the comparison, If you are using a sorting method that has the coin rolling on side you could just have a camera on each side checking for the pattern in the right spot watching at least one rotation of the coin, that way you would only have to check one spot seeing that the coin will be at the desired orientation somewhere within the range... cheap scanning tech, keep the software simple enough and maybe the collectors that have deeper pockets would buy at a premium.

 

Cool! I'll buy one!

 

Calm down, this posting is from 2013

 

[QUOTEgood by [/QUOTE]
You should edit your post and remove your personal information.

 

He is a lawyer. He probably wants publicity.

 

This type of solicitation is not allowed by members and was removed. Please read the rules. this forum is for numismatics, not for negotiating business deals. Thanks, Jim

 

Does that mean you aren't interested in that real estate deal we talked about

 

Then you should also remove the version in the quote

 

Done

 

It takes the fun away of looking at each coin, and doesn't take into account errors.
The date alone is not enough.
Seems that this is just an aid for CRHer's.

 

PCGS has been working on a machine since 1991 which primary function would be to grade coins. I haven't found any actual reports that it is operational, but I highly suspect they use it to grade mass marketed HSN coins and such. I've already written an inquiry, but I don't anticipate a straight forward response.

The patent has been updated recently. Thus it can be concluded they are still pumping money into this project (25 years of building and research adds up to a lot of money). From what I can tell, with a couple slight modifications, this machine could be used for pulling dates and varieties. I don't think it's there quite yet due to technological limitations, but it will be in due time.

 

Maybe IBM can get Watson interested in coin collecting. He needs a hobby.

 

Don't you think that this would ruin the hobby for the little guys?

 

I'd reckon that depends on how you look at it. On one hand, it would make more varieties available, lowering the costs. However, I assume you are saying that it would make coin-roll hunting obsolete. It could lower the numbers of desirable coins in circulation across the board, but even the fastest machine could only sort a limited number of coins per day. Besides, computerized optics are still a long long way from being as efficient as the human eye. At a certain point, if the numbers of desired coins are reduced enough, it wouldn't even be cost-effective to run a machine (the law of diminishing returns). Finally, there's almost always new stock entering circulation from change jars and other stashes. For example, the technology to efficiently remove silver from circulating coinage has been around for a while now, but CRH's still find nominal amounts while hunting.

 

I'm going to have to disagree on that point. For constrained tasks, machine vision can be MUCH more efficient (faster, more accurate, more repeatable) than human vision.

Excellent point. Discriminating silver is really easy -- CoinStar and self-checkout machines routinely reject silver coins. But you'll still find them in the reject slots from day to day, even after the 2011 silver run-up and all the silver scavenging that that drove.

 

Definitely more repeatable (ADD ). More accurate.. is a bit debatable. It depends on what we consider for accuracy since a computer can still only do what it is programed to do. If an anomaly arises which it doesn't have a programmed response for it cannot react. A human has the ability to adapt to random variables at a moments notice.
(for example: on some graded reverse proof coins there are small spots in the fields which don't have the "cameo" effect. To the human eye these stand out like a red flag. ..but apparently they snuck past the grading company and were given a 70. I've seen these most often in First Strike designated holders which i have come to assume are graded with their machine.)

Computerized optics has a huge advantage in fringe areas which are outside of the human eye's limitations, but I don't see them replacing standard visual inspections anytime soon. The cost to perform these type operations by computer starts adding up very quickly due to huge consumption of energy and set-up costs.

fwiw-Not arguing. I understand what you are saying and I don't even really disagree. I think we are just looking at it from different angles. It starts getting into a very knitty analysis of what we consider efficient in practical application. I don't really anticipate this type of tech to be cost effective until true AI exists even as the optics themselves improve. (also... I just find this type of topic fascinating- I really hope one of the photo experts will come around with a direct analysis of current tech vs. human abilities)


oh.. and one more point I forgot in my past post. In high volume machining one of the largest problems is performing operations without damaging the product. Right now, in order to pump coins quickly thru the high-volume sorting machines, they warsh them (it looks like with some sort of detergent) so the gunk doesn't clog up the machine. We also tend to see coins which have roller damage around the edges (in the past wheel marks). So, doing this type of thing quickly and efficiently without damaging the coins is also a huge hurdle.

 

I've also been working on this. I thought I could OCR photos. Then realized that would not work. Resize then match but what about the bright or dark. Then I thought
I could use a 3d scan to measure the height of the
ridges. Later I realized my sensors could only read cm. Which was way to big I'd need nano meters or something.
Speed at which this would work is a small issue. However
accuracy is highly important. Measuring the surface of the
coin would be critical. I will keep experimenting. One thing I thought of out of scope was a chip (sipo) so i could open
and close gates for a drop off collector. So i could fill rolls by date. Either way I wish to know your progress.

 

He hasn't been on for 4 1/2 years, so Probably he has moved to something else.