AI POS Suspicious Activity Detection for Self-Service Checkout

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The increasing use of self-service checkout methods has led to a higher likelihood of shoplifting, necessitating a more effective means to detect suspicious activities in retail environments.

Innovation Solution

A POS system equipped with cameras, a server, and a terminal that utilizes a computer model to analyze facial expressions, appearances, and actions of customers, generating warning messages for store clerks when suspicious activities are detected.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If self-service checkout methods are increased, then productivity is improved, but shoplifting increases

Engineering Contradiction:
Improvecheckout speedVSAvoidshoplifting
Core Design Contradiction:
ProductivityVSObject-affected harmful factors

Solution Approach 1:

The system performs preliminary detection of suspicious behaviors through AI-based image analysis before shoplifting occurs. The computer model analyzes facial expressions, appearances, and actions in real-time to identify potential shoplifters, enabling preventive action before the harmful act is completed.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent introduces an intermediary AI-based detection system between the customer and the shoplifting act. This intermediary system (camera + computer model) monitors customer behavior and provides warnings to store clerks, serving as a mediator that prevents shoplifting without directly interfering with normal customer transactions.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Measurement precision

If AI-based detection system is implemented, then detection precision is improved, but device complexity increases

Engineering Contradiction:
Improvedetection accuracyVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system uses a computer model that has been trained using images and text data to create a virtual representation of customer behavior patterns. This copied knowledge allows the system to detect suspicious activities without requiring complex real-time analysis hardware, as the detection logic is pre-encoded in the trained model.

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The patent replaces complex mechanical surveillance systems with an AI-based software solution. Instead of using multiple cameras, sensors, and mechanical detection devices, the system uses a computer model that processes simple image inputs to generate detection outputs, substituting physical complexity with computational intelligence.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

3Speed

If real-time image analysis is performed, then response speed is improved, but energy consumption increases

Engineering Contradiction:
Improveresponse speedVSAvoidenergy consumption
Core Design Contradiction:
SpeedVSUse of energy by moving object

Solution Approach 1:

The computer model is pre-trained using extensive image and text data before actual detection operations. This preliminary training allows the model to make rapid, accurate detections in real-time without requiring continuous high-energy computation during actual use, as the heavy lifting is done during the training phase rather than during operation.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP4597454A1Point-of-sale system, server, and storage medium
Publication Date: 2025.08.06 TOSHIBA TEC KK
  • EP4597454A1 patent drawingFigure 1
  • EP4597454A1 patent drawingFigure 2
  • EP4597454A1 patent drawingFigure 3

AI summary

A point-of-sale system installed in a store includes a camera installed in the store, a terminal operated by a store clerk in the store, and a server including a network interface and a processor configured to: acquire an image captured by the camera, input the image to a computer model and generate using the computer model in response to the input of the acquired image first text data indicating one of a facial expression, an appearance, and an action of a customer in the image, determine whether the customer in the image is engaging in a suspicious activity based on the first text data, upon determining that the customer is engaging in a suspicious activity, generate a warning message indicating said one of the facial expression, the appearance, and the action of the customer, and control the network interface to transmit the generated warning message to the terminal.