Product Scanner Glare Detection for Item Substitution Fraud
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing solutions for detecting item substitution fraud at Self-Service Checkout (SSCO) and Point-Of-Sale (POS) terminals are inefficient due to reliance on extensive databases and image-based approaches that are prone to errors caused by specular reflection, leading to unreliable results and increased hardware costs.
Innovation Solution
Implementing a product scanner system that processes images using down-sampling, background subtraction, and pixel-by-pixel comparison, along with algorithms such as silhouette detection, glare detection, and Hough transform to identify features and differentiate between man-made and produce items without relying on item databases, enabling real-time detection of item substitution fraud.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If image-based approaches are used for product identification, then the system can detect item substitution fraud, but specular reflection causes unreliable results
Solution Approach 1:
The patent converts the harmful effect of specular reflection into a useful feature by detecting glare patterns. Instead of treating glare as noise to be eliminated, the system uses glare detection algorithms to identify reflected light patterns that help distinguish between produce items (which typically don't have strong specular reflections) and non-produce items (which often have packaging with glare). This transforms the harmful reflection into a beneficial detection cue.
2Measurement precision
If extensive item databases and reference product images are used, then product identification accuracy improves, but hardware costs and computational resources increase
Solution Approach 1:
The patent extracts only the essential features needed for fraud detection - specifically silhouette shape, size, and glare patterns - rather than using complete product images and extensive databases. By taking out only the critical identification elements, the system achieves sufficient accuracy for detecting item substitution while dramatically reducing database requirements and computational complexity.
Solution Approach 2:
The system performs partial image processing by focusing only on key features (silhouette, size, glare) rather than analyzing complete product images. This partial action approach provides sufficient information for fraud detection without requiring full product identification capabilities, thereby reducing hardware and computational requirements.
3Reliability
If texture-based produce classification techniques are used, then produce verification is attempted, but reliable results are not achieved
Solution Approach 1:
The patent changes the detection parameters from texture-based analysis to shape-based (silhouette) and optical property-based (glare detection) analysis. By shifting from texture features to silhouette and glare parameters, the system achieves more reliable produce verification since these features are more consistent and less susceptible to the variations that plague texture-based methods.
4Measurement precision
If multiple cameras are used for image capture, then detection accuracy improves, but processing time increases
Solution Approach 1:
The system performs preliminary processing by capturing multiple images simultaneously or in rapid succession, then immediately processes them to identify the best quality image for analysis. This preliminary action of capturing multiple images upfront allows the system to achieve high detection accuracy through image selection and processing, while maintaining fast overall processing time by avoiding repeated captures.
Data Source
AI summary
Various embodiments herein each include at least one of systems, methods, software, and devices, such as product scanners e.g., barcode scanners) enabled to detect item substitution fraud during a checkout process at facilities, such as retail outlets. One method embodiment that may be performed in part on a product scanner, includes detecting, on a point-of-sale (POS) terminal, an event occurrence associated with at least one validation process. The method may then receive an image from each of at least one camera of a plurality of cameras of the product scanner. The method further includes processing at least one of the received images according to the at least one validation process to obtain a result and providing the result to the POS terminal.