Case study

Metrici AI quality check in bank industry


Brink's is a security company specialized in high values transportation and fully management of cash flow for both companies and bank industry. It helps companies reduce costs, have access to cash and automize internal processes. It is the biggest CIT company in the world and also provides full service of ATM manipulation.

The processing and values deposit which Brink's is offering reffers to cash manipulation (RON and some other), and also deposit for cash, gold, stocks etc.


The company is one the leaders for maintenance and reload of ATM machines in Romania. It is a daily routine for thousands of such ATM cassettes all around the country as much of the population is massively still using cash for everyday transactions. The activity of loading the ATM cassetes is done in company’s main headqurters in different parts of the country.

All the process of reloading ATM cassette is made by human operators and although mistakes are somehow rare, they still occur. So the client wanted to eliminate them all together. True to its promise to process and deposit values for its clients in the best conditions and for optimizing the costs, the company needed a solution to help them deliver. Our trusted partner, Eutron, works with Brink's offering technical support and approached Metrici for its AI capabilities and its performance in recognizing different objects and situations.

Eutron is a romanian company offering personalized technological solutions for over 30 years. They are focused on helping partners by improving the experience of their clients, optimize their cash flow, offer security solutions or data identification and space management. Over time, company's experience was shared with over 3000 clients in retail, finance, health industry, production or logistics, one of which is Brink's itself. As such, the collaboration between Eutron and Metrici was a natural process and a normal step to take and jump in to help Brink's.

A special DNN training was made for recognition of multiple objects for a number exceeding 31 classes: all Romanian denominations, the cassette themselves, the lids for that cassettes, different type of labels for each denomination and also, euro banknotes and their designated labels.

An IP camera is used for each ATM cassette loading station with one operator. All cameras are connected to a central server, where analysis is done by Metrici detection engines.

Metrici detects when the operator begins a new loading round by recognizing the cassette and its lid. It reads a special label to know which kind of denomination must be loaded in that particular cassette.

As the operator takes banknotes and load them into the cassette, Metrici keeps track of the activity. If a wrong type is loaded (comparing the marks on the cassette with the banknotes from operators hands or which are already placed in), Metrici will trigger an alarm, both audio as well as visual: red light on a Metrici MultiController.

An event is recorded into the database with this error, containing information about what station was, timestamp, photo.

Technical sheet for each product line/ location

  • 1 camera for each loading docks (usually up to 18 per location )
  • 1 Metrici detection engine for each loading dock
  • Dedicated AI training for recognizing over 31 objects classes
  • A HA/LB cluster of two servers working together for best performance and availability
  • 1 Metrici MultiController with led lights
  • 1 sound alarm

How the system works

No single move inside the detection area can be made without Metrici noticing it. Always on the guard, it will sound an alarm every time a mistake is made- namely wrong banknotes charged in the ATM cassette.

The banknotes go into cassettes based on their denomination. Metrici acknowledges the cash bundles, now accurately assesses and scrutinizes the environment which ensures a high level of accuracy.
In such a way, the replenishment is done only supervised by Metrici, which offers non-stop monitoring.

This comes along a high level of transparency, with real-time data capturing, updates and notifications, quick and accurate analysis, investigation and resolution of possible errors. The system is to be expanded to all company's locations and DNN training are to be pursued for the other denominations that our partners are working with.