Currency exchange is one of the most common operations in banks. The customers come in to exchange their money regularly and the banks need to do the exchange and register the operations in the banking systems to provide a secure service to the customers and keep the internal banking records up-to-date.

In reality, it is more complicated. Some cashiers take advantage of the situation. They don't register the exchange operations and simply put the difference in the pocket. It is a small fraud that implies big reputational risks to the banks, since a cashier who started with pocketing small amounts of money may later switch to bigger schemes, getting into the clients' personal data and accounts. These incidents are also hard to detect since the bank security service needs to process manually hours of video records from all branches, which will be time-consuming and will involve the human factor.

Vision Systems has created an intelligent video analytics solution based on Artificial Intelligence and Computer Vision. The system analyzes the videos from the security cameras at the bank cash counter in real-time and matches the events in the videos with the data in the bank system to detect fraudulent activities with the currency.

The system has allowed the bank to keep the currency exchange operations under control. After a few months of using our solution, the bank has managed to detect several fraudulent operations and dismiss the violators, thus preventing financial and reputational losses. With the growth of the currency exchange operations and the opening of new branches, it will be hard to underestimate the effect that the software will have on bank operations.

If a customer has put the banknotes for exchange in the cash tray, the system detects them and recognizes the currency. Then it registers the currency exchange operation and connects with the automated bank system to check whether the cashier has registered the operation. If he hasn't, the system notifies the security service about the incident. It is worth mentioning that our system can recognize banknotes even if they have been partially covered with IDs, papers, and credit cards.

The system also detects that a previous customer has left and the new one has come in and a new exchange operation has started. This feature may be upgraded in future versions of the solution and allow verifying the faces of the customers based on the person's ID.

Going forward the solution can be trained to recognize more currencies to prevent fraud in a wider range. Also, additional features can be added to lessen the workload of the cashiers: the solution can help autofill the personal data on the receipt given to the customer based on the data in the ID.