ISJ hears exclusive insights from Dr. Igor Kudryashov, CEO, Quantum Detection.
Retailers worldwide continue to lose billions of dollars each year to what the industry calls shrinkage – a combination of shoplifting, employee theft, administrative mistakes and supplier fraud.
Recent studies estimate losses between 1.5% and 2% of total sales, translating into tens of billions annually. For major retail chains, that can mean the difference between profit and loss.
Traditional loss-prevention tools – CCTV cameras, patrols and manual post-incident reviews – were designed for a slower era.
They rely on human vigilance, which is limited by fatigue and distraction.
Even with high-definition video, valuable evidence is buried in thousands of hours of footage, usually reviewed only after an incident occurs.
Meanwhile, the modern retail environment has grown far more complex.
Self-checkout, omnichannel logistics and reduced staff on the floor have opened new vulnerabilities. Thieves are more organised and insider theft is harder to detect when operations span hundreds of locations.
Across industries, this challenge is driving a new generation of AI-based human detection and behavior-analysis systems – technologies capable of recognising people and their activities across multiple cameras and detecting suspicious or abnormal behavior in real time.
Such systems are already finding applications beyond retail, from critical infrastructure and transportation hubs to ports and border checkpoints, where they help identify unauthorized access or unsafe behavior.
Quantum Detection, a global developer of AI software for X-ray and CCTV analytics, brings experience from airports, logistics hubs and border inspection systems into the retail sector.
The company’s expertise in object recognition, human tracking and behavioural analysis has enabled it to create a flexible platform that not only enhances security but also improves operational visibility.
In this article, we focus on how the same technology – applied intelligently – can dramatically reduce losses in retail, transforming ordinary surveillance networks into proactive, data-driven security tools.
AI is transforming the way surveillance systems operate. Instead of simply recording video, modern AI solutions analyse live camera streams, identifying people, objects and behavioural patterns that may indicate potential risks.
The result is a shift from passive observation to active prevention, where systems can detect and alert before a loss occurs.
In retail, one of the key technological challenges is person re-identification.
Most stores are already equipped with a network of CCTV cameras positioned at different angles, with varying resolutions and lighting conditions.
Recognising the same individual as they move between aisles, entrances and checkout areas requires advanced neural algorithms capable of compensating for these differences.
Since facial recognition is increasingly restricted or forbidden in public spaces under privacy regulations such as GDPR and CCPA, re-identification has become both a technical and ethical challenge – AI must learn to track individuals without relying on biometric identifiers.
Quantum Detection has addressed this challenge through proprietary algorithms originally developed for airports, ports and border checkpoints, where both accuracy and compliance are critical.
Adapted for the retail sector, this technology enables cross-camera tracking and behavioural analysis while respecting privacy boundaries.
The system can identify abnormal movement patterns – such as repeated shelf visits, concealment attempts or unusual checkout behavior – and correlate them with contextual data like point-of-sale events or staff access logs.
Rather than replacing human security, AI augments it, delivering real-time insights that help prevent incidents before they happen and turning existing CCTV networks into proactive protection systems.
Behind every alert or insight lies a set of deep-learning algorithms that analyse movement, posture and interaction rather than appearance.
Instead of recognising faces, Quantum Detection’s system identifies up to 17 anatomical key points for each person, capturing pose, gait and hand position across different camera views.
By doing so, it builds a dynamic model of human behaviour – how someone walks, reaches for a product or conceals an item – without ever storing biometric data.
In addition to people, the system simultaneously tracks shopping carts, baskets and bags, allowing it to understand context: Whether a product is being legitimately placed into a cart or bagged prematurely or if a trolley is left unattended in a sensitive area.
These correlations between human and object movement provide a clearer behavioural picture and dramatically improve accuracy.
A further layer of the re-identification process is outfit detection.
Clothing colour, texture and silhouette help maintain consistent identification across different camera angles and lighting conditions – crucial in retail environments where faces may be obscured or where privacy laws prohibit facial recognition.
The system then combines pose, gait and outfit features to follow individuals as they move through the store, even when switching camera zones.
Some of the images accompanying this article illustrate how the AI successfully detects people from various angles and camera types, including wide-angle and fisheye lenses, where distortion and perspective changes can challenge traditional analytics.
The ability to maintain detection quality under such conditions demonstrates the robustness of the technology and its readiness for real-world retail environments.
Processing can be done on-premises or at the edge, minimising bandwidth usage and ensuring privacy compliance. Integration with existing CCTV networks, access systems and POS terminals requires no hardware replacement – only a software layer that turns ordinary video into actionable intelligence.
The result is an always-learning network that sees beyond pixels and objects.
It understands how people, their belongings and their behaviour evolve over time, enabling retailers to prevent theft, optimise staff deployment and enhance customer experiences – using the cameras they already own.
The age of reactive video review is ending.
AI-driven systems like those developed by Quantum Detection are redefining how retailers understand what happens in their stores, turning everyday cameras into intelligent, collaborative observers that reduce losses, increase staff awareness and protect profitability.
By combining behavior analysis, re-identification and contextual object tracking, loss prevention becomes not just about stopping theft – it becomes a data-driven process that strengthens operations and customer experience alike.
As retail continues to evolve, those who embrace intelligent vision today will set the standards for tomorrow’s security and efficiency.