Self-Checkout Action Determination for Product-Personal Item Distinction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional self-checkout systems struggle to accurately distinguish between products and personal possessions, leading to incorrect bar-code reading determinations due to the similarity in appearance of various items and personal items like wallets or smartphones.
Innovation Solution
An action determination program that acquires user operation history on the self-checkout machine, specifies actions involving objects grasped by users, and determines whether the action involves a product registered for purchase by analyzing image data and history information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If conventional object recognition technology (e.g., YOLO) is used to detect products, then detection speed is improved, but accuracy deteriorates due to inability to distinguish products from personal possessions
Solution Approach 1:
The detection process is segmented into two distinct stages: first, conventional object recognition (YOLO) rapidly identifies all objects in the scanning region; second, a determination unit analyzes each detected object to classify it as either a product or personal possession based on additional criteria. This segmentation allows the system to maintain high detection speed while improving accuracy through specialized analysis in the second stage.
Solution Approach 2:
A determination unit is introduced as an intermediary component between the object recognition system and the final detection output. This intermediary analyzes characteristics of detected objects (such as movement patterns, positional relationships with the checkout machine, and temporal behavior) to accurately distinguish products from personal possessions, thereby resolving the accuracy issue without compromising the speed of the underlying object recognition system.
2Reliability
If the system detects all objects in the scanning region, then detection completeness is improved, but false positive rate increases due to confusion between products and personal items
Solution Approach 1:
The system employs dynamic analysis of object behavior and characteristics rather than static detection. It monitors movement patterns, temporal presence in the scanning region, and interaction with the checkout machine to dynamically determine whether an object is a product being registered or personal possession being moved. This dynamic approach maintains detection completeness while reducing false positives through behavioral analysis.
Solution Approach 2:
The determination unit changes the parameters used for object classification from simple visual recognition to multi-dimensional analysis including movement velocity, trajectory, temporal duration in scanning region, and spatial relationship with the checkout machine. By changing these parameters, the system accurately distinguishes products from personal possessions, maintaining completeness while reducing false positives.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
An action determination program includes acquiring history information of a user operation on a checkout machine to which a product to be purchased is registered and that performs checkout processing of the registered product, specifying, based on an image obtained by capturing a user in front of the checkout machine, an action in which the user operates the checkout machine while grasping an object, and determining, based on the acquired history information, whether the action in which the user operates the checkout machine while grasping an object is an action in which a product to be purchased is registered to the checkout machine.