Self-Checkout Fraud Detection with Camera and POS Event Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Self-checkout systems face challenges in accurately distinguishing between fraudulent and non-fraudulent acts, leading to false accusations and customer complaints due to the limitations of AI image analysis.
Innovation Solution
A detection device integrated with a camera and processor that analyzes both camera images and point-of-sale events to identify potential fraudulent acts, setting thresholds for reporting such acts, and notifying both the customer and a supervisor when the threshold is exceeded.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If AI image analysis is used to detect fraudulent acts, then the ability to detect fraudulent acts is improved, but false detections increase
Solution Approach 1:
The detection system is segmented into multiple independent analysis modules: AI image analysis module, POS event analysis module, and integrated judgment module. Each module processes specific types of data independently, and their results are combined through the integrated judgment module to reduce false detections while maintaining fraud detection accuracy.
Solution Approach 2:
The integrated judgment module acts as an intermediary between the AI image analysis and POS event analysis. It receives detection results from both modules, performs cross-validation, and makes the final determination, thereby reducing false positives from either individual module.
2Reliability
If AI image analysis is used to detect fraudulent acts, then fraud detection capability is improved, but customer complaints increase
Solution Approach 1:
The system implements feedback mechanisms where detection results are validated against multiple data sources before triggering notifications. The integrated judgment module continuously refines detection accuracy based on the correlation between image analysis results and POS event data, reducing false accusations and subsequent customer complaints.
Solution Approach 2:
The system performs preliminary cross-validation of detection results before issuing fraud alerts. By pre-checking AI detection results against POS event data and applying integrated judgment criteria, the system prevents false accusations from reaching customers, thereby reducing complaints.
3Measurement precision
If multiple detection methods are integrated, then detection accuracy is improved, but system complexity increases
Solution Approach 1:
The complex detection system is divided into modular components: image acquisition module, POS event acquisition module, AI analysis module, and integrated judgment module. Each module has a specific function and can be independently developed, tested, and maintained, reducing the practical complexity despite the advanced detection capabilities.
4Measurement precision
If threshold-based reporting is implemented, then false accusations are reduced, but some fraudulent acts may be missed
Solution Approach 1:
The system dynamically adjusts detection thresholds based on the integration of multiple data sources. The integrated judgment module modifies the effective threshold by considering both image analysis confidence levels and POS event correlation strength, ensuring that legitimate fraud cases are not missed while maintaining low false accusation rates.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
According to one embodiment, a detection device for a registration system has a communication interface connectable to a point-of-sale terminal, a camera, and a monitoring terminal. A processor of the detection device receives a camera image from the camera depicting a customer at the point-of-sale terminal and point-of-sale event signals from the point-of-sale terminal. The point-of-sale event signals correspond to operations of the customer performed at the point-of-sale terminal. The processor determines, based on the received camera image and the received point-of-sale event signals, whether a reportable act has occurred at the point-of-sale terminal and outputs a report of the reportable act to the point-of-sale terminal if a threshold number of occurrences set for the reportable act has been met during the operations of the customer at the point-of-sale terminal. The report causes the point-of-sale terminal to provide notice to the customer.