Barcode-Aware Object Verification for Retail Fraud Prevention
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current barcode verification systems are vulnerable to 'ticket switching' fraud, where a barcode from a lower-value item is swapped with a higher-value item, leading to incorrect pricing during transactions, due to the inability to accurately associate barcodes with their corresponding objects, especially under varying capture conditions such as different orientations and illumination.
Innovation Solution
A barcode-aware object verification system that uses a machine learning-based classifier and item templates to extract a consistent region-of-interest from captured images, matching the barcode with the object by comparing the barcode's location, orientation, and size to the template data, and generating a sub-image for classification, thereby ensuring accurate object identification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional barcode scanning is used, then scanning speed is fast, but accuracy of barcode-object association deteriorates due to ticket switching fraud
Solution Approach 1:
The patent introduces an item template as an intermediary data structure that bridges the barcode and the actual item. The template contains pre-stored geometric and contextual relationships between the barcode and item features, allowing the system to verify associations without direct complex analysis of every captured image, thus improving accuracy while managing complexity.
Solution Approach 2:
The system performs preliminary actions by pre-defining item templates with expected geometric relationships between barcodes and item features before verification occurs. This allows the verification process to simply compare captured images against predetermined templates rather than performing complex real-time analysis, improving both accuracy and efficiency.
2Adaptability or versatility
If image processing is performed under varying capture conditions, then adaptability improves, but measurement precision deteriorates due to orientation and illumination variations
Solution Approach 1:
The patent applies parameter changes by transforming the captured image to match the orientation and illumination conditions of the stored item template. This includes rotating, scaling, and adjusting brightness/contrast parameters of the captured image so that geometric relationships can be accurately compared despite varying capture conditions, thereby maintaining measurement precision while achieving adaptability.
Solution Approach 2:
The system creates equipotential conditions by normalizing the captured image to the same geometric and photometric reference frame as the stored template. This eliminates the effect of varying capture conditions by transforming all images to a common reference state, allowing accurate comparison regardless of original orientation or illumination.
3Loss of information
If the entire item image is processed, then completeness is improved, but processing time increases due to large image areas
Solution Approach 1:
The patent extracts only the relevant region-of-interest from the captured image based on the barcode location and the geometric relationships defined in the item template. This extraction focuses processing on only the necessary portion of the image that contains the item features needed for verification, reducing processing time while maintaining completeness of essential information through the template-guided selection.
Data Source
AI summary
A method captures first image data representing at least a portion of a first item and decodes a first barcode in the first image data and determines a first item template associated with the first barcode. The first item template comprises first identifier data identifying the first item from other items and first region-of-interest data specifying a first region-of-interest of the first item. Based on the first region-of-interest, a second region-of-interest of the first image data is determined. The second region-of-interest is modified to fit within a boundary of the first image data. The method determines whether a ratio of an area of the second and first regions-of-interest is greater than a first threshold and generates second image data comprising the second region-of-interest when the ratio is greater than the first threshold. The method determines the second image data corresponds to the first identifier data and generates first data indicating the first barcode is matched with the first item.