Dynamic Detection Line Adjustment for Retail Product Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional camera-based product recognition systems for self-checkout and shoplifting detection face challenges due to varying positional relationships between cameras and shelves, leading to erroneous product detection, as the predefined detection lines do not adapt to different user positions and imaging conditions.
Innovation Solution
A computer-readable recording medium stores a program that acquires video from cameras in a store, uses machine learning to identify depth and generate a 3D in-store model, sets detection lines based on aisle ranges and directions, and adjusts these lines dynamically according to user positions, enabling accurate detection of product taking in and out.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If a detection line is uniformly defined in advance for camera-based product recognition, then the system is easy to implement and install, but detection accuracy deteriorates due to varying positional relationships between camera and shelf
Solution Approach 1:
The patent transforms the static detection line into a dynamic one that adapts to each user's position. The detection line is no longer uniformly defined in advance but is dynamically determined based on the user's detected position relative to the shelf, allowing the system to maintain high detection accuracy across different positions while still being easy to implement
Solution Approach 2:
The patent changes the parameters of the detection line based on user position. Instead of using a fixed detection line, the system adjusts the detection line's position and orientation parameters according to where the user is standing, enabling accurate detection for each user while maintaining system simplicity
2Device complexity
If a fixed detection line is used for all user positions, then the system complexity is reduced, but detection reliability deteriorates due to erroneous detections
Solution Approach 1:
The patent applies local quality by setting different detection lines for different user positions. Each position has its own optimized detection line that is specifically suited for that location, improving detection reliability without requiring a completely complex system architecture
3Measurement precision
If the detection line is adjusted for each user position, then detection accuracy is improved, but the computational complexity and processing time increase
Solution Approach 1:
The patent performs preliminary action by pre-establishing the relationship between user positions and detection lines. The system detects user position first, then quickly determines the appropriate detection line based on this position information, reducing processing time while maintaining high detection accuracy
Data Source
AI summary
A non-transitory computer-readable recording medium has stored therein a setting program that causes a computer to execute a process, the process including acquiring a video from a camera, identifying a depth indicating a distance from the camera to each of constituent elements of the video acquired from the camera, generating a three-dimensional in-store model, generating skeleton information on a person who moves inside the store from the video acquired from the camera, setting a range and a direction of an aisle in the store in the generated three-dimensional in-store model based on a change in the generated skeleton information and setting a detection line in the storage based on the range and the direction of the aisle in the store, the detection line for detecting that the person has extended a hand to a product.