Indicia Reader Fraud Detection via Image Embeddings
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Self-checkout devices lack a reliable method to detect fraudulent activities, such as rescan attempts, without the need for bulky and expensive weight scales, which increases security risks for retailers.
Innovation Solution
A method using an indicia reader with a controller and memory to detect objects, generate image embeddings, and compare them to support samples, triggering alerts or preventing transactions if a match is not found, thereby identifying potential fraudulent activities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If weight scales are used to detect fraudulent activities at self-checkout devices, then security reliability is improved, but device size and cost increase
Solution Approach 1:
The patent replaces the mechanical weight scale system with an optical/image-based detection system. The indicia reader captures images of objects, generates embeddings from these images, and compares them to detect fraudulent activities. This substitution eliminates the need for bulky mechanical scales while maintaining security detection capability.
Solution Approach 2:
The patent creates digital copies (embeddings) of object images to represent the physical objects. Instead of measuring physical weight, the system captures visual information, converts it to embedding representations, and uses these digital copies for comparison and fraud detection. This copying approach reduces device complexity while preserving reliability.
2Measurement precision
If image data is captured and processed to generate embeddings for fraud detection, then security accuracy is improved, but computational time and processing complexity increase
Solution Approach 1:
The patent performs preliminary actions by capturing images and generating embeddings during the normal scanning process, before fraud detection is needed. The embeddings are created in advance and stored, so when fraud detection is required, the comparison can be performed quickly without intensive real-time processing. This preliminary preparation reduces the time loss during actual detection.
Solution Approach 2:
The patent transforms image data into embedding representations, changing the parameter space from raw pixel data to compressed feature vectors. This parameter transformation reduces the dimensionality and complexity of the data, enabling faster comparison and processing while maintaining or improving detection accuracy through the use of meaningful feature embeddings.
Data Source
AI summary
A method and system for operating an indicia reader are disclosed herein. An example method includes detecting a first object in the scanning region; capturing one or more images of the first object to create first image data; determining that a successful decode of an indicia has not occurred; retrieving the first image data of the first object; generating, using the first image data, an embedding of the first image data; detecting a second object in the scanning region; retrieving one or more support samples from an image database; comparing the embedding of the image data to each support sample of the one or more support samples; based on the comparison that the embedding of the image data does not match any support sample, performing an operation.