Skeleton-Based Information Processing for Self-Scan Fraud Alerts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems for self-scanning in retail environments struggle to accurately detect fraudulent behaviors, such as scan omissions, due to varying product importance and attention levels, leading to potential delays or missed responses, especially when dealing with large numbers of undetected items.
Innovation Solution
A fraud detection system that utilizes a fraud detection device connected to cameras and user terminals, employing object and skeleton detection techniques to evaluate customer behavior, assign priority levels based on product importance and customer attributes, and notify store clerks of fraudulent activities through a store clerk terminal.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a monitoring camera system is used to detect fraudulent behaviors in self-scanning retail environments, then the coverage and detection capability are improved, but the accuracy of detecting scan omissions deteriorates due to varying product importance and attention levels
Solution Approach 1:
The system assigns different priority levels to different commodity products based on their importance and attention levels. High-priority products receive more intensive monitoring and analysis, while low-priority products receive standard monitoring. This localized quality adjustment resolves the contradiction by concentrating detection resources where they are most needed, improving overall detection accuracy without requiring uniform high-intensity monitoring across all products.
Solution Approach 2:
The system dynamically changes detection parameters such as analysis depth, notification thresholds, and monitoring intensity based on product priority levels. For high-priority products, the system applies stricter detection parameters and more comprehensive analysis, while for low-priority products, it uses standard parameters. This parameter adaptation resolves the contradiction by optimizing detection accuracy for each product category rather than applying a one-size-fits-all approach.
2Reliability
If all commodity products are monitored with equal attention in self-scanning systems, then the detection coverage is improved, but the response time and efficiency deteriorate due to the large number of undetected items
Solution Approach 1:
The system segments commodity products into different priority categories (high, medium, low) based on their importance and attention levels. This segmentation allows the system to apply different monitoring and response strategies to different product groups. By dividing the monitoring task into segmented priority levels, the system maintains comprehensive coverage while improving response efficiency for critical products, resolving the contradiction between coverage and efficiency.
Solution Approach 2:
The system applies partial monitoring intensity to different product categories rather than excessive uniform monitoring. For high-priority products, it applies excessive monitoring to ensure no scan omissions occur, while for low-priority products, it applies standard or reduced monitoring. This partial action approach maintains adequate coverage across all products while concentrating resources on high-priority items, thereby improving overall response efficiency.
3Reliability
If the system notifies store clerks of all detected fraudulent behaviors, then the detection completeness is improved, but the information processing load and false positives increase
Solution Approach 1:
The notification system applies different notification strategies to different priority levels. For high-priority scan omissions, the system generates immediate and detailed notifications to store clerks. For medium-priority items, it uses standard notifications, and for low-priority items, it may use aggregated or delayed notifications. This localized notification approach maintains detection completeness while reducing the overall notification load and false positives, resolving the contradiction between completeness and system complexity.
Solution Approach 2:
The system changes notification parameters such as threshold values, aggregation levels, and alert priorities based on the detected fraud severity and product importance. For significant scan omissions involving high-priority products, it uses low thresholds and immediate notification. For minor violations involving low-priority products, it uses higher thresholds and aggregated notification. This parameter adaptation reduces false positives and notification complexity while maintaining detection completeness.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An information processing program causes a computer to execute a process including: generating, by inputting a captured image into a machine learning model, skeleton information on a person included in the captured image; detecting, by using the skeleton information, a specific motion of the person related to an object included in the captured image; specifying, by using positional information on the person included in the captured image, a first area in which the person is located at a time of detection of the specific motion from among a plurality of areas; specifying, by reading setting information recorded in a memory, first setting information that is associated with the first area; and identifying, based on the first setting information, a priority level of a notification of an alert related to the specific motion of the person related to the object.