This is my first post – been reading the forum for a few weeks and joined about a week ago. Anyway here is what I’ve been working on – actually it is only a concept at this time because computer programming is not my forte. Thought it over and decided to post – looking for suggestions, input and maybe even help in developing this idea. I have posted a picture (I hope) for reference purposes to aid in explaining what I am trying to accomplish. My concept is based on the following theory. Mint marks prior to 1989 were hand punched into working dies. The ability of a person to duplicate this punch in alignment, orientation and location from one die to the next is highly unlikely. Being that all US coins are easily attributable to date and mint, duplicating the same punch within the same year is even more remote. Therefore I have concluded that every die which carries a hand punched mint mark is unique and a coin struck by that die carries a fingerprint (so to speak) which can be attributed to a specific working die utilizing the coin’s type, date, type of mint mark (P, D, S, etc.) and coordinates of the mint mark in relation to the overall design of the coin. The picture I have attached with this explanation is a Lincoln Cent but the principles can be applied to any coin struck by a working die containing a hand punched mintmark. You select 3 plotable devices in the overall design of a coin series and utilize these points as standard benchmarks for analyzing every coin of that series. The Lincoln Cent I have chosen Benchmark 1. The point where the part in Lincoln’s lips come together. Benchmark 2. The point in B of Liberty where the curves converge. Benchmark 3. The point in W of IGWT where the inside lines converges. The next step is plotting many points around the perimeter of the mintmark. The photograph I attached has the benchmarks constructed into a triangle (for illustration purposes I believe the triangle makes it easier to visualize the process). So – how about a piece of software in which you acquire an image of a coin from a flatbed scanner. You plot the standardized benchmarks and mintmark. The image is resized and oriented proportionally based on the distance between the benchmarks. The location and orientation of the mint mark relative to the benchmarks is located using triangulation of the points plotted around the mint mark and the results are compared to a known RPM, Double Die and etc database. Sort of a GPS (Loran) fingerprint approach to die variety attribution. What do you think? Have I been watching too much CSI? Harebrained? Doable? Any computer programmers out there? Oh; by the way – if anyone steals this idea and develops this software or anything similar – I expect a free full version working copy.
Howdy justafarmer - Welcome to the Forum !! An intersting idea - one that's been in use for many, many years - but it's not done by a computer. But I suppose it could be. You'd have to program in allowances though, seldom is a coin struck exactly centered between the dies and not all planchets are exactly the same size. And each fraction of a millimeter in difference would matter with a program such as you describe. I think the reason nobody has done this is because it is easier to just learn the die diagnostics and recognize the coins by looking at them. Many older coins are collected by die pair.
Yes it would work and I think it's a great idea. I was kicking it around myself. Wouldn't it be great to lay 30 coins on a scanner and within minutes know exactly what you have. The program itself is easy and would consist of two parts. The first is the scan in which the coin image is brought in and croped out of the main scan. The second part is the recoginition. Mind you all of this is speculation but as I said I have thought about it. The next step of the program would need to be able to determine the type of coin by determining the rim distance or diameter. Using a contrasting identification method this should be relatively simple. The next step would be to use the contrasting identifier to mark out the image outline. Once the outline is formed the image is checked against the master file to determine the reverse or obverse image it matches. If no match is made the immage is rotated a specified amount and the matching process would be repeated. This is done over and over until the images orientation is determined. Once determined it can easily be compared to a database of marks since all of the other areas should be consistant. This explination can obviously be extrapolated further but in theory it should be relatively straight forward for a programer who is familiar with imageing sofware cod to create. I have to assume that the scanner would have to be higher than most run of the home units but a company such as HP may be intrested enough to cost share a research proposal to fund the project.
Very good question,about every mintmark being unique. Something I've done is to take pictures of coins with marks that are tilted or otherwise odd looking,go to a photo program and put the coins side by side and correct as to the degree of rotation, blow them up,print them out and then use an inclineometer or just a steel rule to compare.
Centering of the strike wouldn't make any difference. Since triangulation is based on benchmark points in the design devices. These remain constant being an inherent property of the working die striking the coin.
off topic Just noticed the G.Forks location, my g.dad & dad are from there, and just had visitors from there who left here just today. How highs the water?
Well two days ago I would have said low and in the banks but we got 5+inch this morning and we are looking at some more tonight. They've been issuing flash flood warnings all day. Could see the river out of the banks by morning and mabey close to flood stage within two days.
Well I hope your on high ground,and keep your socks dry. I think my forebearers left there because of the floods.
Well that depends on what your benchmark points are I suppose. But also keep in mind - there can be literally thousands of working dies for each coin series & date - and each one of them is going to be different.
True but if your just comparing to known error varities there are usually only a few recognized per year.
Don't think the quality of the scanner would be a big issue. The program wouldn't even use the actual image to make its analysis. The scanned image would be utilized by the user for the purpose of plotting the points on an overlay. The overlay is what the computer analysis would be based.
GDJMSP wrote: "Well that depends on what your benchmark points are I suppose. But also keep in mind - there can be literally thousands of working dies for each coin series & date - and each one of them is going to be different." That is really my point and the basis of my concept. Each die is different but every coin struck by that die inherits the GPS fingerprint of the devices of that die. In my theory every working die with a handpunched mintmark could be identified and attributed. Whether it is a collectable die variety or not would be denoted in the database.
