Self-Checkout Fraud Detection via Weight Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Self-checkout machines face challenges in detecting fraud and errors, such as force majeure errors and intentional fraud, which can result in unpaid amounts due to scan omissions or barcode concealment, making it difficult to accurately calculate product weights and prices, especially for items like fresh foods.
Innovation Solution
An information processing device that uses machine learning models to estimate product weights by acquiring video data and weight information from cameras and accounting machines, generating training data to train models that detect fraudulent behavior by correlating product attributes with weight information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If self-checkout machines are introduced to reduce labor costs, then labor cost is reduced, but fraud detection capability deteriorates
Solution Approach 1:
A weight estimation system acts as an intermediary between the self-checkout machine and fraud detection. The system includes a camera to capture product images, a machine learning model to estimate weight from images, and a comparison module to detect discrepancies between estimated and actual weights, enabling fraud detection without additional manual labor
Solution Approach 2:
The patent replaces manual weight verification (mechanical/human process) with an automated machine learning-based weight estimation system. The system uses image processing and machine learning models to automatically estimate product weights and detect fraud, substituting the need for human intervention while maintaining detection capability
2Measurement precision
If traditional weight measurement is used for fresh foods, then weight measurement is straightforward, but fraud detection accuracy deteriorates due to scan omissions and barcode concealment
Solution Approach 1:
The system implements feedback by comparing the machine learning-estimated weight with the actual weight measured by the scale. When discrepancies exceed a threshold, the system flags potential fraud. This feedback mechanism enables continuous verification and improves fraud detection accuracy by cross-validating weight information through multiple independent measurement methods
Data Source
AI summary
A non-transitory computer-readable storage medium storing an information processing program that causes at least one computer to execute a process, the process includes acquiring video data of a product placed on a scale included in a registration machine; specifying an attribute regarding an appearance of the product, by inputting the acquired video data into a first machine learning model; acquiring information regarding a weight of the product, from the accounting machine that has measured a weight of the product placed on the scale; and performing machine learning of a second machine learning model, by using the specified attribute of the product as training data and the acquired information regarding the weight of the product as correct answer data.