scanning & sorting machine for baseball cards

[tallan]

I work with business who deal in millions of loose-leaf baseball cards. They currently pay hundreds of thousands of dollars each year to have part time workers count & sort these cards for resale at locations across the US.

With recent advances in OCR technology and mass-market scanning products (like business-card readers and cash-counting machines) I am searching for an already-existing solution that can be modified for this task rather than spending $500,000 for all-new custom manufacturing.

The two biggest problem areas (mechanical) seem to be:

1) Condition sensitive: If a machine is going to 'grab' a card repeatedly to read/sort/place it, how to minimize damage; as even the slightest scratch on a card can sometimes drop value by $1000+

2) The (geometrical) sorting bin/shelf concept. i.e. once a card is identified, how to most-effiently 'put it away' amongst a million other already pre-sorted cards. Should there be 'rows of cards' in 100ft-long bins and a machine hand runs along the top picking & pulling? Or should the cards be kept in shorter bin/drawers, like a card-catalog in a library? and so on. And finally, when cards are all stacked up in a row together, they are difficult to 'pull' out, i.e. to pull out just one card without pulling (or damaging) its neighbors.

Also chain-linking (via usb or similar) several smaller machines together may have efficiency gains over a single massive machine. Undoubtedly the smaller the machine the better.

This task already has a budget set aside for it, so any ideas feel free to contact me directly, whether individual engineer or full corporate. I would especially be interested in hearing from anyone who has already built similar machines (xerox copiers, coupon sorters etc).

 

[MintJulep]

I would start with the companies that make paper money handling equipment for vending machines.

 

[imagitec]

Presumable, they will be known - unique identification, value, condition - before entering the system. They will be stored in a known location, and that index location will be stored in a database with the other identifying information so that it can be retrieved later by the same mechanism.

To minimize wear and tear & by extension loss in value, you'd want to store each in an isolated compartment so it could be placed and retrieved without touching any other. But with hundreds of thousands of such compartments (I'm assuming that the millions of cards they deal in represents turn over and not inventory), that would be cost and space prohibitive.

However, you know the value of each unique card, so you don't need to treat them the same. A card worth $1000 would be stored in a single card silo. A 10 cent card would be packed tightly with hundreds of other similarly priced cards. The 10 cent cards would see more disturbance with greater risk of damage, but even significant damage would have minimal price impact, as opposed to slight damage causing a $1000 drop in value of an expensive card, which you cited.

I can imagine the cheap cards standing vertically, front-to-back, clamped between a stationary bookend a moveable bookend. Move the moveable book end slightly and tilt it forward and the cards will fan out. Count the cards from the fixed end to locate the one you want. Always insert new cards at the moveable end.

I can imagine the valuable cards in individual trays stacked one on top of each other, like a desktop in-out box. The trays would have a cutout to permit a vacuum end effector to move under it, pick it up in a single vertical motion (don't want relative lateral movement between the end effector and the card, which could result in scratches) and carry it away.

 

[GregLocock]

so far as the bit you haven't talked about goes, check out the computer vision toolbox for matlab from mathworks. It may do everything you need for the image reecognition side.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[tallan]

some good thoughts here so far, I will follow them up thanks

 

[imagitec]

Another backend consideration: Sort and store the cards in a way that distributes access to bins - and therefore minimizes wear and tear - somewhat equally. Here's a hint: It won't be alphabetically. I was reminded to this as I waited to vote on primary day. My ward had separated the voter checklists into three lines: A-I, J-R, and S-Z. I wished I was a Zanini that day.

Random would be a decent method.

If historical access data is available, a smarter alogrithm could be developed.

Rob Campbell
Imagitec: Imagination-Expertise-Execution