Checkout Video Tracking for Self-Scan Fraud Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing self-service checkout systems face challenges in detecting fraud, such as scan omissions and bar code errors, due to the high cost of weight sensors and the difficulty in training image recognition AI for diverse and rapidly changing product types.
Innovation Solution
A system using a camera and machine learning model to track hand and product movements, identifying abnormal behaviors by analyzing video images and generating alerts for fraudulent actions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If weight sensors are installed to detect fraud, then fraud detection accuracy is improved, but device cost and complexity increase excessively
Solution Approach 1:
The patent replaces mechanical weight sensors with a vision-based detection system using cameras and image processing algorithms. The system captures video images of the checkout process and uses computer vision to track product movement, hand movements, and detect anomalies such as scan omissions or barcode hiding, thereby eliminating the need for expensive mechanical weighing equipment while maintaining fraud detection capability
Solution Approach 2:
The system creates a visual copy of the physical checkout process through video imaging. By capturing and analyzing images of the scanning area, product movement, and user actions, the system replicates the functionality of weight-based detection through optical means, allowing fraud detection without physical contact or additional sensing hardware
2Reliability
If image recognition AI is trained to detect fraud, then detection capability is improved, but training difficulty increases due to diverse and rapidly changing product types
Solution Approach 1:
The patent segments the fraud detection task into distinct components: product detection, hand detection, movement tracking, and anomaly classification. By dividing the complex image recognition problem into smaller, specialized sub-tasks, the system can train separate models for each function, reducing the overall training difficulty and improving adaptability to different product types
Solution Approach 2:
The system changes the approach from training AI to recognize specific product types to detecting generic patterns of fraudulent behavior. Instead of learning product-specific features, the AI focuses on movement patterns, spatial relationships, and temporal sequences that characterize fraud, making the system adaptable to any product type without retraining
3Productivity
If self-service checkout is implemented to reduce labor costs, then labor efficiency is improved, but fraud detection capability deteriorates due to user self-operation
Solution Approach 1:
The patent implements continuous visual monitoring and real-time feedback during the self-service checkout process. The system analyzes video streams to detect anomalies such as products not being scanned, barcode hiding, or unusual hand movements, and can provide immediate alerts or warnings to users, maintaining fraud detection capability while preserving self-service operation
Solution Approach 2:
The system enables the checkout process to monitor and detect its own anomalies through automated visual analysis. The AI-powered vision system acts as a self-monitoring mechanism that independently identifies fraudulent behaviors without requiring additional human operators, allowing self-service to remain self-regulating
Data Source
AI summary
An information processing apparatus acquires video image data on a person who is scanning a code of a commodity product to an accounting machine, and specifies, from the acquired video image data by analyzing the acquired video image data, a region of a hand of the person and a region of the commodity product that is being gripped in the hand of the person. The information processing apparatus tracks either a movement of the hand of the person that is gripping the commodity product, or, a movement of the gripped commodity product, and generates, based on a change in the tracked movement of the hand or a change in the tracked movement of the commodity product, an alert connected to an abnormality of a behavior of registering the commodity product to the accounting machine.