Home
Cashier Fraud Monitoring
Automatic detection of self-checkout fraud by AI,
Reduce store losses and achieve cost reductions by addressing labor shortages.
Reduce store losses and achieve cost reductions by addressing labor shortages.
Challenges in Retail
Retailers face the dual challenge of labor shortages and cost reduction.
While self-checkout systems are increasingly common, they've led to a rise in fraud like missed scans and "banana tricks," resulting in significant revenue loss.
Traditional monitoring methods are inadequate, exacerbated by rising labor costs and persistent staffing issues.
While self-checkout systems are increasingly common, they've led to a rise in fraud like missed scans and "banana tricks," resulting in significant revenue loss.
Traditional monitoring methods are inadequate, exacerbated by rising labor costs and persistent staffing issues.
Our Solution
Our AI and image processing technology automatically detects fraudulent activity at self-checkout.
The AI analyzes camera footage and cross-references it with POS data to identify missed scans, barcode switching, and other irregularities in real-time.
This reduces monitoring burdens and achieves cost savings by addressing labor shortages.
The AI analyzes camera footage and cross-references it with POS data to identify missed scans, barcode switching, and other irregularities in real-time.
This reduces monitoring burdens and achieves cost savings by addressing labor shortages.
Value Proposition
- Highly Accurate Fraud Detection
- Accurately detects various fraudulent activities, minimizing revenue loss.
- Reduced Labor Costs
- Automates employee monitoring tasks, reducing labor costs.
- Deterrent Effect
- The system's presence acts as a deterrent, reducing the incidence of fraud.
- Easy Implementation
- Requires only standard cameras and a GPU-equipped PC, enabling faster deployment.
|
|
|
|
|
|
|
|
|
|
|
|
Demo
App
Click here to try the demo app
Technical Overview
Target Industry / Users
Retail industry, self-checkout managers
Challenges in Target Industry & Operations
While self-checkout adoption is increasing as a solution to chronic labor shortages and cost reduction,
the following issues remain, making self-checkout fraud prevention a key challenge:
- Increasing store losses due to accidental or intentional fraudulent activities at self-checkout.
- Traditional reliance on employee monitoring is unsustainable due to labor and cost constraints.
Technical Challenges
The goal is to automatically detect fraudulent self-checkout activity using AI.
However, addressing the diverse nature of fraudulent activities remains a challenge, including:
- Scan avoidance：Items intentionally or unintentionally left unscanned.
- Banana trick: Selecting a cheaper item on the screen for an unscanned item.
- Label switching: Covering a high-priced item's barcode with a lower-priced one.
- Remained unscanned items：Leaving unscanned items in the shopping basket before completing checkout.
Solutions
Automatic detection and notification of self-checkout fraud through video analysis-based object and person recognition.
- Kozuchi-integrated features:
- Scan avoidance detection:
Counts items using video analysis and compares this with the number of scanned items from the POS system to automatically detect missed scans.
- Scan avoidance detection:
- Kozuchi-independent features (reference):
- Detection of Scanned item falsification (banana trick, label switching):
Recognizes items using video analysis and compares this with the scanned item information from the POS system to automatically detect incorrect scanned (registered in the POS system) items. - Detection of remained unscanned items in baskets
Detects remaining items in the shopping basket using video analysis and determines if the basket is empty when checkout begins to automatically detect abandoned items.
- Detection of Scanned item falsification (banana trick, label switching):
Fujitsu's Technological Advantage
- Robust video analysis technology for detecting and recognizing items brought to self-checkout.
- World’s most accurate detection of “An object held in hand”. (2023)
※Detection accuracy for “An object held in hand” on HICO-DET dataset (presented in VISAPP 2023) - 1st prize in competition of product recognition at self-checkout on international conference. (2023)
※CVPR AI City Challenge 2023
- World’s most accurate detection of “An object held in hand”. (2023)
The benefits of Cashier Fraud Monitoring (Detailed version)
- Automatically detects various fraudulent activities at self-checkout, reducing the burden on stores while minimizing losses. The system's presence also acts as a deterrent, preventing further losses.
Use Cases
- End-users: Retail store operators
- By installing one general-purpose camera per self-checkout kiosk and an edge PC with this technology installed, automatic fraud detection is possible for multiple self-checkout kiosks. Real-time notifications to staff and backend log review functionality are also provided.
- App developers
- Install this technology on an edge PC using an installation package and train the necessary "item detection AI" and "item recognition AI" for video analysis in the customer's environment, enabling rapid environment setup.
Case Studies
- Numerous proof-of-concept tests(PoC) with target industry clients.
Technical Trial
- Demo app: Try the web app
- PoC available
- Concept Video
- Demo Video
Related Information
- Fujitsu TECH BLOG
Documents
| Document Name | Description |
|---|---|
| Instructions(PDF) | Instructions for the Cashier Fraud Monitoring demo app. |
| License Information(PDF) | License information for OSS used in the Cashier Fraud Monitoring demo app. |
Contact Us
The Cashier Fraud Monitoring demo app only showcases "Scan avoidance detection" as part of the overall technology. For details on the complete technology, please contact:
- Email：fj-tsdx-ml-qa@dl.jp.fujitsu.com
Demo
App
Click here to try the demo app