CCTV enhanced with algorithms and AI: how can Artificial Intelligence be used to enhance security in retail stores?
The retail sector is particularly affected by shoplifting. Since having security staff is not always enough to catch suspicious individuals, algorithm-enabled video surveillance is becoming a real asset as far as a store’s security is concerned. But what makes this solution so effective?
Algorithms, AI, algorithm-enabled video surveillance: a few definitions
Artificial Intelligence is a field of science that’s constantly evolving. Since one of its purposes is to imitate human intelligence, AI can be found in a wide range of industrial and business applications, including video surveillance.
Artificial Intelligence and algorithms
The term Artificial Intelligence covers all types of simulated human intelligence. AI is computer software developed based on a variety of theories and techniques. AI has a wide range of applications: smart personal assistants, chatbots, decision support systems, image recognition, and home automation, to name but a few. In particular, by using algorithms, AI can be used to solve complex problems.
In the truest sense of the term, an algorithm is a sequence of clearly defined and detailed operations that can be used to solve a problem. An algorithm is also a computer-executable program that uses a certain amount of data to perform a task.
For this, developers use a process called machine learning. This involves training the computer (the machine) by providing it with a large amount of data.
Definition of algorithm-enabled video surveillance (AVS)
Algorithm-enabled video surveillance is one of the applications of Artificial Intelligence. It combines video surveillance with algorithms to increase its effectiveness. The aim here is to create an autonomous program capable of detecting suspicious behaviour or suspicious-looking shapes, especially with a view to combatting shoplifting.
A problem common to all stores: the explosion of shoplifting incidents
Shoplifting is not a new phenomenon. In order to protect themselves against this growing scourge, many retailers equip their stores with security cameras. Despite this measure, retailers have seen a sharp increase in this type of crime over the last few months. Shoplifting can occur in any type of store on a daily basis, including pharmacies, grocery and other food stores, and even hardware stores.
More incidents due to inflation
Inflation seems to be largely responsible for the upsurge in shoplifting. Dwindling purchasing power is prompting people to resort to stealing. New shoplifter profiles have emerged. Even the most loyal customers can be tempted to steal, whatever their age.
Food aisles are particularly hard hit
According to reports in the press and other media, shoplifting seems to mainly affect food products. Many store managers are now putting security tags on food products.
Online reselling of stolen goods
Stolen items are also being resold online. From smartphones to DIY items, some shoplifters will try to steal and resell anything. Buyers of stolen goods don’t always think to ask for the original invoice.
Monopolised security resources
As a result of this upsurge in shoplifting, a store’s organisation can suffer. In addition to the loss of revenue, this scourge is forcing security staff to redouble their vigilance. Security guards spend an enormous amount of time dealing with individuals they suspect to be shoplifters. Sometimes it’s already too late, and other times it’s simply a mistake and the individual has done nothing wrong. Whatever the outcome, it’s a waste of the security guard’s time.
AVS developments and trends
Algorithm-enabled video surveillance is not a new application. It has already been used for a number of years now for facial recognition. But a number of other uses have also been developed, particularly in the field of security. With AI, it’s now possible to use security cameras to protect stores against shoplifting while at the same time respecting the privacy of shoppers.
Developing AVS to enhance store security
Developers have focused their research on the real-time analysis of security camera footage. This data is used to train the software to detect suspicious behaviour (i.e. through machine learning). The data can be used to create an algorithm capable of learning autonomously and analysing more complex data (deep learning) and thereby improve efficiency and performance.
Veesion software is based on this very same algorithm, which is capable of detecting suspicious gestures or behaviour in any size of store in real time.
Veesion software can be installed on any type of video surveillance software kit and on security cameras already installed in the store. The software connects to the store’s video recorder and synchronises with the existing cameras. Don’t forget: our AI technology can even be installed in a store with only one camera.
Video alerts sent to your smartphone or tablet
Once the software is installed, the first alerts are sent within 30 minutes upon detection of suspicious behaviour. These notifications are basically short videos. They are sent in real time to a smartphone or tablet, for example.
A customisable tool
Veesion software can be configured to suit the needs of the store, taking account of opening hours, staff numbers, and the number of cameras, for example. It can be set to operate round the clock, both day and night. The tool’s return on investment can be measured thanks to weekly and/or monthly reports. These reports may include statistics on the number of shoplifters caught, or the value of goods recovered.
Protection of sensitive data
The software only collects essential data, securing it in compliance with French law and European regulations. All store-related information, as well as the personal data of employees and customers is protected. No data is shared with a third party, or monetised in any way whatsoever.
The benefits of in-store AVS
In the past, shoplifting mainly affected aisles such as alcoholic beverages, cosmetics, hygiene and clothing. However, owing to the high inflation we have seen in recent months, there has also been an increase in food theft.
Store managers are turning to anti-theft devices such as security tags, gates and cameras to improve the security of their stores. But in doing so, retailers also need to hire staff to monitor security screens. What’s more, the human eye is not infallible. The benefits of algorithm-enhanced video surveillance are therefore obvious.
Your security camera video stream monitored in real time without let-up
AVS analyses video images and footage in real time. This means it offers immediacy. As soon as an anomaly, i.e. suspicious behaviour, is detected, a video alert is immediately sent to the person in charge of security. This makes it possible to take action before the suspected individual leaves the store. Any shoplifter is literally caught red-handed.
Greater detection of suspicious body movements
Veesion enables security staff to spot more suspicious behaviour. Certain body gestures and movements can be difficult for the human eye to perceive or spot, while others may go undetected or escape the attention of guards, especially at the end of a shift. Security guards can now rely on video alerts to catch shoplifters.
Fewer false alarms
Algorithm-enabled video surveillance actively reduces the number of false alarms, since it can distinguish a suspicious body movement from an innocent one. AVS focuses exclusively on critical events. This means that security staff never have to worry about wrongly intercepting a customer who is merely shopping. What’s more, video alerts enable security guards to decide whether or not to take action. If the AI does happen to issue a false alarm, it automatically learns from its mistake.
Optimised security resources
AVS can be set up to operate round the clock, 24/7. So no need to have a team of security guards on hand to monitor all the different security screens, entrances and exits. Video alerts enable greater availability. Security staff can therefore be assigned to other shifts and/or tasks, and rotated more easily.
AVS challenges and concerns
Artificial Intelligence often arouses intrigue, and is not without its concerns or questions, particularly with respect to its ultimate aim and purpose. It is sometimes criticised for being an invasion of privacy and considered unethical. Does algorithm-enabled video surveillance really invade people’s privacy? Are we moving towards mass surveillance? Is there a risk of error or algorithm-based bias?
AVS and privacy can go hand-in-hand
Veesion is committed to developing ethically responsible technology.
- works exclusively by analysing body gestures and movements that appear suspicious. The software does not recognise faces, nor does it analyse a person’s emotions, gait or dress;
- does not allow stores to track a shopper or any other individual inside the store;
- does not record the identity of an individual, whether suspected of or actually caught shoplifting;
- Is a decision-making tool: it’s always the human being who ultimately decides whether or not to challenge a suspected shoplifter.
To ensure legal compliance, we recommend you consult the CNIL website. The CNIL (or French National Commission for Data Protection and Liberties) is an independent administrative authority that provides information on how to go about installing security cameras and how to process the personal data collected by those cameras.
Real-time analysis of body movements: some examples
What type of behaviour does the AI algorithm look for? What kinds of body gestures and movements might be considered suspicious? Here is a non-exhaustive list of examples that we used to train our Veesion software:
- Slipping an item into a pocket or garment;
- Concealing a product in the bottom of a pushchair;
- Not scanning all items at the checkout;
- Opening a product while in the store;
- Consuming food while in the store.
Near-zero risk of error
Because Veesion software uses automatic learning and deep learning, any errors in interpreting body gestures are always only minimal. It has been trained to recognise suspicious behaviour by imputing millions of examples of specific body movements and gestures associated with shoplifting. And since it never stops learning autonomously, it can only ever get more effective.