The Retailer Summer Edition 2021

THE RE TA I L ER

20

MACHINE VISION – IMPROVING OPERATIONAL EFFICIENCY AND MAINTAINING BRAND STANDARDS

David Nicholls CTO Retail & Hospitality Fujitsu

M achine vision makes the store of the future, the store of now. The new Amazon Go stores demonstrate technology can deliver a convenient expe- rience for customers and at the same time reduce operating costs. Many retailers facing a similar challenge of demanding customers and squeezed margins will be watching how this venture progresses closely. Most might not be ready to dive-in with fully automated stores just yet due to cost or practicalities like size of store, but they could benefit from one of the key technologies behind Amazon Go. Machine Vision. Machine Vision applies artificial intelligence to visual data, captured by cameras that could range from your existing CCTV systems, to specialised heat sensing cameras that can spot variations in fridge temperatures or electrical cabling at distance. Recent advances in this technologymean that if it can be seen by camera, in most cases it can now be recognised bymachine. This turns cameraswithmachine vision into ‘Omni-sensors’ removing the need for multiple store technol- ogies, such as thermometers, gate controls and footfall counters. For a while, algorithms have been able to tell us not just that a person has walked in, but also their gender, and approxi- mate age. Today the technology has advanced considerably so not onlywill it recognise phys- ical characteristics, but it will also identify over 100 different behaviours. Want to knowwhen someone is getting agitated and there may be a risk of violence? Want to identify a mobil- ity-disabled customer who might need help? Want to spot an item going into a coat rather than a basket?Want to be certain promotions are displayed correctly? All these can be iden- tified by Machine Vision. Fujitsu’s technology can even ensure that staff are completing all six stages of handwashing. Even though hands come in different sizes, and the movements are complex, subtle and obscured with soap. The recognition of detailed observations will make Machine Vision an indispensable tool for delivering operational efficiency and brand standards for retailers.

Colleague Safety The BRC Crime Survey found there were 450 incidents of abuse and violence towards shop staff every day. No surprise then that 90% of retailers say colleague safety is a top three concern for the board. A challenge for store colleagues facing an agitated member of the public is that remaining calm is essential to de-escalate the situation, but at the same time signalling for help can be seen as the very oppo- site of this. Machine Vision could help here, as it is able to spot or even predict an escalating situation. It could even spot pre-agreed signals from colleagues to raise a silent alarm or even an automated announcement warning of risk of prosecution. This makes it clear that the situation has been spotted, but in a way that draws focus and hopefully danger away from the colleague.

Better and Faster SCO experience

With each Self-checkout unit on average losing over £11,000 a yearwith non-scans many retail- ers would be glad of an extra set of eyes here. Machine vision can spot non-scans and product, bar code replacements and immediately offer customers the opportunity to re-scan or fix the error. This automation can avoid a potential flashpoint with a colleague. Profit protection isn’t the only way Machine Vision can benefit the SCO process. Product recognition can cut scanning times for improve- ments in productivity and reducing basket abandonment. Fujitsu has already deployed MachineVision at SCOwith two major French supermarkets to classify non-barcoded items such as produce and bakery and offer custom- ers a ‘Top 3’ dynamic menu with an accuracy of 97.66%.

In 2021 Fujitsu successfully achieved the world’s highest accuracy against the world standard

benchmark in the field of behaviour recognition.