Store Video Interaction Detection for Self-Checkout Alerts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing self-checkout systems struggle to analyze the depth relationship between bounding boxes in two-dimensional space, making it difficult to identify interactions between persons and objects, and they lack effective means to prevent shoplifting.
Innovation Solution
A system using a machine learning model to analyze video data from cameras, identify regions of products and persons, and detect anomalous motions, such as grabbing, to output alerts when predetermined conditions are met, thereby preventing shoplifting.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a machine learning model is used to extract bounding boxes from video data, then object detection capability is improved, but the ability to analyze depth relationships and identify interactions between persons and objects deteriorates due to two-dimensional spatial limitations
Solution Approach 1:
The patent transitions from two-dimensional bounding box coordinates to three-dimensional spatial relationships by calculating depth distances between persons and objects using coordinate differences. This dimensional enhancement enables the system to analyze spatial relationships and interactions that were previously undetectable in 2D space, resolving the contradiction between detection capability and depth analysis ability.
2Productivity
If self-checkout cash registers are introduced to improve labor efficiency, then productivity is improved, but the ability to prevent shoplifting deteriorates due to lack of monitoring
Solution Approach 1:
The system implements continuous feedback by analyzing video data in real-time to detect anomalous behaviors such as shoplifting. The machine learning model processes video feeds, identifies suspicious interactions between persons and objects, and triggers alerts to store personnel. This feedback mechanism maintains high productivity through automated checkout while simultaneously restoring shoplifting prevention capabilities through intelligent monitoring.
3Reliability
If automated video analysis is implemented to detect anomalous behaviors, then shoplifting prevention is improved, but system complexity increases due to machine learning model requirements
Solution Approach 1:
The machine learning model serves multiple functions: it extracts bounding boxes from video data, calculates spatial relationships between detected objects, identifies interactions between persons and products, and detects anomalous behaviors. This multi-functionality reduces the need for separate specialized systems, thereby improving shoplifting prevention while limiting the increase in overall system complexity.
Data Source
AI summary
A non-transitory computer-readable recording medium stores an alert output program for causing a computer to execute a process including: acquiring a video of an inside of a store where products are arranged; specifying, from the acquired video, a first region that includes the products, a second region that includes a person of interest who is to purchase the products, and a relationship in which an interaction between the products and the person is identified, by analyzing the video; determining, when the relationship satisfies a predetermined condition, whether or not a motion carried out with respect to the products included in the first region by the person included in the second region has an anomaly, based on the relationship; and outputting an alert that indicates that a person who makes an anomalous motion with respect to the products has appeared, when the motion is determined to have the anomaly.