Building sensing model- currency

[pfbrady@virginmedia.com]

I sold a car this weekend, that means I have an unusual situation of having a 55 typical used bank notes- lucky me, but not for long, they wont be around for long!

It would be nice to use the Scio as a counterfeit note detector

My question is about building the model. When I create the model I want to scan all the notes in  3 different locations giving me different reference points/scanning locations for any given note. This means that for any note tested it can be checked in three possible places, particularly as people often write on banknotes.

The basic attributes are ” Note value”£20, Scan location- “Rear plain area”, “Queens head”, “front hologram” and the final attribute is the surface under the note in this case “A4 plain paper”.

I have a choice on building the model and as I don’t know how Scio handles the difference.

Option 1- the easy one- do this for each scanning location.

Stack the notes up and scan them one by one in the same location on the notes and then repeat the whole stack probably 3 times for each scanning location. Then I have three “samples”, one for each scanning location with multiple scans for each location done on different notes.

Option 2- The really tedious way

Stack the notes up and for each scanning location create a new sample for each note and scan the same note in the scanning location multiple times. This method involves creating a new “sample” for each note which takes time between scans.

 

My question is-I will gather the same number of scans with identical attributes by both methods, will scio handle it differently if it is “organised” differently  eg Option one multiple scans with different notes categorised as the same sample. or option 2 multiple scans on different notes categorised as different samples.- Confused?

 

Anyway this is a small-ish model with 55 odd notes scanned multiple times it’s a very important question for much larger models, it dictates whether it takes days or hours to collect the samples. Some guidelines on the most efficient method of building a model would be helpful.

 

Hope  this makes sense

Paul

Attachments:

You must be logged in to view attached files.

[rejsharp]

Hi Paul;

I think option 1 where you scan the same note and same location multiple times as a new sample is wrong – it is NOT a new sample, and would unfairly tighten the spread of results. (Multiple scanning within a single sample ID is good).

Making the model simple will help speed of scanning – stick to white paper under the note as standard. (I would test whether a single sheet is enough, the illumination does penetrate somewhat).

You may need to have Condition as an attribute, with New, Used, Grubby as values.

Given the complex patterns on a note, how will you cope with slight inaccuracies in the area illuminated? Will you use a jig? Otherwise that variance may be greater that that of bad notes.

Also – how do you hope to recognise forgeries? Do you have access to some bad notes? or will you try to detect differences from expected scans? i.e. outliers. I would not know how to do that, but maybe CP can offer good advice!

[pfbrady@virginmedia.com]

Thanks REJSHARP,

I suspect this will only every be a “toy” application, however it’s useful in building my experience for more serious matters. Thanks for your reply and your points are excellent. I intend to get some flat sheet PTFE and use it to make a jig to introduce repeatability. I have a background that includes quality management so Gauge Reliability and repeatability (gauge R&R) is something I understand. I thought I would share the scans to date just for giggles. Two show a useful spectrum. One is the rear of the note which is mostly white so it’s a measure of the paper. The queen’s head is mostly purple ink with an embedded watermark. Both of these would seem a reasonable basis for a model. The third spectrum is over the holograms-hehehe. Not exactly repeatable….

Have you been up to anything interesting with yours?

Attachments:

You must be logged in to view attached files.

[Guy]

Hi Paul (and Roger)

Roger is right – method 2 is the right way to create your data collection. SCiO’s algorithmic engine requires DIFFERENT samples to have a successful model. Scanning all notes under the same sample id will result in a poor model. To write it differently – 100 scans of a single sample is not the same as 10 scans of 10 samples.

 

Using PTFE as a reflective surface is a good idea – it does not have prominent feature in NIR and is supposed to be very reflective.

 

I would also scan notes of a different denomination to see if there’s a difference between them, and obviously, if you have access to counterfeited notes, it will be very good to scan them as well.

 

Keep us posted 

 

 

[Scott]

It would seem much more likely that this would be (as you say) a toy application — showing notes which have been in contact with cocaine, notes smeared with disease-ridden kid snot, etc. 🙂

Also, the Bank of England has instructions on non- or low-technology techniques to identify counterfeit bills.
http://www.bankofengland.co.uk/banknotes/Pages/retailers/default.aspx

Instructions are also available for Euro currency:
http://www.new-euro-banknotes.eu/Euro-Banknotes/Security-features/LOOK/THE-NEW-%E2%82%AC20

and US currency:
https://uscurrency.gov/

Given this, I don’t see an app as more than toy, as you say, although I’m sure you can find lots of interesting facts, such as identifying US bills which have been tested using a marker. I suspect you’re unlikely to see a real counterfeit bill — I don’t see any police department agreeing to let you test known counterfeits; they’ll probably refer you to the treasury.

–Scott.

[Scott]

Here’s a thought for you: allowing functionally-blind people (I recently read a novel about a completely blind/deaf/mute ex-cop) with poor touch (calloused or damaged fingertips?) and/or with worn bills to identify bills. I know Euros are different sizes, and I don’t know about UK bills, but US bills are all the same size. Perhaps an app to help blind people be more certain that they haven’t been handed a fake, or handed a lower denomination than they were told. You probably wouldn’t need to include voice or morse-code-vibrate capability; there should be apps which handle that already.

With the computer revolution, most clerks (except in banks) no longer count money out loud by denomination; they just silently drag out a wad of bills, causally riffle though it to be sure it’s in the ballpark (why do you say I sound cynical??), and give it to the customer.

Yes, I know plastic is probably a better overall solution for a blind person…

–Scott, who is hoping his eyes will never get that bad.

[Author]

Posts