How Computer Vision Technology Can Bring Smart Surveillance to Retail    

Even after the surge of online shopping, brick-and-mortar stores around the world are still thriving. However, unlike their virtual counterparts, physical stores continually face the challenging task of preventing shoplifting, customer theft and refund fraud.

Shoplifting costs the retail industry billions of US dollars every year and is a major contributor to inventory shrinkage and loss. Many retailers are boosting budgets for their loss prevention teams, with investments in technology increasing the most.

Retailers Lose Billions from Shoplifting

According to the 28th annual 2019 National Retail Security Survey in the United States, jointly conducted by the National Retail Federation and Dr. Richard Hollinger of the University of Florida, shrink — or loss of inventory due to shoplifting, theft, error, or fraud — has remained stable over the past few years. However, when this average shrink rate of 1.38% is extrapolated to the entire US retail industry, it shows an estimated US$50.6 billion loss for the sector, the survey showed.

A total of 68.2% of surveyed professionals with loss prevention teams at major US retailers responded that they would allocate additional resources for loss prevention, mostly in technology.

Shoplifting has a significant impact on another major retail market, the UK.

The 2019 Retail Crime Survey published by the British Retail Consortium showed that customer theft – theft by customers or persons purporting to be honest customers – directly costs the industry more than £700 million, up by 31% year on year. The total cost of all types of crime for the industry is estimated at £1.9 billion. This is roughly equal to around 20% of the estimated profits of the entire UK retail industry.

Retailers’ Challenges in Spotting Shoplifters

Typically, security staff in physical stores can efficiently process and remember faces. However, there is a limit to the number of faces a person can immediately process, recognize and remember for future reference, especially in the busiest shopping hours and days. In addition, changes in appearance, including beards, wigs, caps or glasses can easily trick a person’s ability to process and compare a face to their memories of images of faces. What’s more, retailers have learned from experience that shoplifters and pickpockets often look like an honest potential customer.

A recent study of a research team from the University of York tried to answer the question ‘How many faces do people know?’. The research suggested that people ‘know’ an average of 5,000 faces, including faces of famous people. But the study found that there is a fundamental difference between identifying previously seen faces and faces never seen before. A person’s ability to identify previously unseen faces can easily be disrupted by a change in image, the authors of the study said.

Computer Vision Helps Efficient Face Recognition

Here is where computer vision technology can help identify and process a huge number of faces and with greater efficiency. The technology is aimed at automating and replicating the cognitive processes of the human visual system. After obtaining information from images and videos, computer vision systems use machine learning methods to train computers to process and analyze patterns across faces.

As a result, retailers are able to spot suspicious persons in their stores, even if people are not looking straight at security cameras or wear glasses, grow beards or change hairstyles.

Ontotext’s Smart Surveillance Solution Using Computer Vision

Ontotext has developed a smart surveillance solution to help retailers with efficient identification and processing of faces in their stores. The solution is based on the computer vision technology that has been developed within Sirma Group since 2013 and now complements Ontotext’s portfolio within Sirma AI.

Ontotext’s smart surveillance system enables retail chains to receive real-time notifications of suspicious persons or behaviors. It creates short video clips and reproduces them retrospectively on demand to show the actions of a person of interest. The solution enables retail chains to keep a central database with images and videos of known offenders and use it in all their stores. This significantly improves the process of efficiently discovering potential offenders and creates a safer environment for both customers and employees.

This smart Surveillance Solution for Efficient Face Recognition was developed for a big Bulgarian chain store.

Are you facing similar problems?