Group Project 12 – Use of Image Processing and AI for detection of a currency note

The idea for this project is based on an RFP floated by the Reserve Bank of India (RBI) to help visually impaired persons in recognizing denominations of currency notes

( https://www.mstcecommerce.com/eprochome/rbi and the website of the RBI at https://www.rbi.org.in under the menu “Tenders”).

How can computer vision technology and AI(ML) be used to help a visually impaired person?

Each note printed by RBI under Mahatma Gandhi Series has unique characteristics; for details please see an example at https://paisaboltahai.rbi.org.in/rupees-two-hundred.aspx.

These features include

• a defined size; each denomination is a different size, larger size for larger value
• defined objects like Mahatma Gandhi portrait, Ashoka Pillar emblem
• an identification mark on the left-hand side of each note on the front side which is in raised print (intaglio) and has different shapes for different denominations for eg.  square for Rs 50, rectangle for Rs 20 and none for Rs 10.
• etc.

The goal of the project is to develop software that would

• recognize correct denomination of an image of a banknote taken by a camera (of a mobile)
• have a very high accuracy of identification
• alert user to properly place/rotate currency note for recognition.

This project can be approached considering:

• segmentation/division of the work on the identification of features among team members and/or over an iteration cycle. Then assemble the feature identification components and have the processing of these components in some logical sequence;
• Processing of feature identification components can be in a layered filter approach i.e. in the first pass let software eliminate an image based on the absence of most common attribute (like size), then progressively go on recognizing the presence/absence of other features;
• Since features to match are well defined, we could also match extracted objects from the camera image with correct features (in stored images of proper currency notes).

Students should attempt a scaled down requirement scope to demonstrate the technology. Scaling down can be based on

• recognition of banknote image of only one denomination
• attempting a recognition of only 2 attributes namely size and presence of one other feature object e.g. Ashoka Pillar emblem
• attempting recognition of proper placement of banknote – is it face up or not ( how do we determine it? – see at the end)
• lastly, seek confirmation from the user if it (image) is a genuine or not to declare the value of denomination.

We need suitable programming languages and algorithms for object identification from captured images and then compare. Recall input from Computer vision/Graphics classes. ML concepts can enhance the efficiency of the identification process.

There are some interesting aspects to note about this project:

• Once we have logic tested for one denomination, it is easy to enhance it for other denominations. Here is a good example of an advantage of OOD/P (which one?).
• Note that we are not attempting feature recognition of a class of items e.g. straight head bolts which have common shape features but different dimensions of length, width. In this project, we are doing one to one matching and hence processing logic could be simpler.

Once we have an acceptable version on desktop, it could serve as a base for a mobile version.

Caution: If a computer vision algorithm identifies the presence of all desired features in a captured image, can we imagine that it represents a genuine banknote? The answer is NO. So what else should happen in our program to increase the accuracy of recognition?

Note:

1. How to find if a banknote is face up or down? Well, there are some features present only on one side of a banknote.

2. How to find if a banknote is folded or note? Well, a comparison of its overall shape will give an idea.

3. How to find if a banknote is genuine or not? Well, Indian banknotes have some features that can be felt by touch. Their presence can be confirmed by a visually impaired person.

References:

1. Currency note design  https://paisaboltahai.rbi.org.in/100.aspx
2. Help for visually challenged on an Indian banknote http://vikaspedia.in/social-welfare/financial-inclusion/know-about-indian-currency/security-features-in-banknotes#section-3