Paper Money

Using computer intelligence to rebuild shredded currency

A PhD candidate at Hong Kong University is developing an idea to use computer vision to reconstruct bank notes from shredded currency.

Images copyright kentannenbaum |

Could those popular little bags of shredded United States currency long sold as novelty souvenirs by the Bureau of Engraving and Printing and available for sale in the Washington, D.C. and Fort Worth visitor centers one day go the way of the dodo?

That is the question posed in a Jan. 21 article by Sue Gee, “Reconstructing Banknotes Using Computer Vision” in the I Programmer journal. The online publication is dedicated to programming and computer technology.

The BEP sells its three-in-one packets of $150 worth of shredded money for $20. Gee writes that in Hong Kong until recently, visitors to the Monetary Authority visitor center could buy a now sold-out paperweight souvenir for $100 Hong Kong dollars that supposedly contained shredded bank notes equivalent to 138 $1,000 HKD bank notes.

A visit to the HKMA prompted Chung To Kong, a PhD candidate at Hong Kong University, to come up with the idea of using computer vision to reconstruct a complete bank note, as if it were a jigsaw puzzle. He has now written a paper, “The possibility of making $138,000 from shredded banknote pieces using computer vision.” The author says it “is intended as a proof of concept so that the banking industry can be aware of this loophole in creating money.” His conclusion: “Use of computer vision has demonstrated a high possibility of reconstructing banknotes from shredded banknote pieces.”

His research focused on the viability of converting shredded bank notes into complete ones, knowing that at least half of the banknote and visible serial number was required. His research was based on a 2011 U.S. Military Defense Advanced Research Projects Agency Shredder Challenge, a team competition with a set of five problems in which the number of documents, subject matter and the method of shredding was varied to present challenges of increasing difficulty. The purpose was to find ways to identify documents destroyed in combat. The $50,000 first prize went to “All Your Shreds Belong to U.S.” The group combined custom-coded computer vision algorithms to suggest fragment pairings, with human verification and assembly. It showed that collaboration between humans and machines could be effective. Kong also said that since 2011, improvements in computer vision and machine learning lessen the need for human participation.

The author also noted that computer vision has also been used to solve jigsaw puzzles. In the case of the Hong Kong bank notes, he said that the irregular shapes and the tiny size of the pieces made it akin to a challenging jigsaw puzzle. The difference is that bank notes have unique serial numbers, and these are printed in two specific locations. Machine learning was used to look for shreds containing parts of serial numbers. Training required only 100 images of two classes — with and without a serial number. Mapping the face and back of bank notes from three different banks required more training, but the conclusion reached was that it could be done.

Connect with Coin World:  
Sign up for our free eNewsletter
Access our Dealer Directory  
Like us on Facebook  
Follow us on X (Twitter)
Keep in touch on MyCollect - the social media platform for collectibles 

Whether you’re a current subscriber or new, you can take advantage of the best offers on magazine subscriptions available in digital, print or both! Whether you want your issue every week or every month, there’s a subscription to meet your needs.

Community Comments