Goods Recognition Verification Using Change Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image recognition systems for managing display shelves in stores have low recognition accuracy, requiring human confirmation and correction, which is labor-intensive and inefficient.
Innovation Solution
An information processing system that includes a displayed goods recognition processing unit and a verification target extraction processing unit to identify and extract recognition results that require verification by comparing current and previous recognition data, using changes in goods information and product tag information to determine verification targets.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If image recognition processing is used to identify displayed goods, then automation is improved, but recognition accuracy deteriorates
Solution Approach 1:
The verification process is segmented into two stages: initial automatic recognition by the image recognition unit, followed by selective verification only for extracted candidates. This segmentation maintains automation while improving accuracy by focusing human or system verification resources on specific uncertain cases rather than all recognized goods.
Solution Approach 2:
The verification target extraction unit acts as an intermediary between the image recognition unit and the final verification process. It filters and identifies candidate goods that require verification, serving as a bridge that reduces the burden on the final verification stage while maintaining the benefits of initial automated recognition.
2Reliability
If all recognition results are verified by human confirmation, then reliability is improved, but productivity deteriorates
Solution Approach 1:
Instead of verifying all recognition results (excessive action), the system applies verification selectively only to extracted candidate goods that meet specific criteria (partial action). This partial verification approach maintains sufficient reliability for critical cases while significantly improving productivity by avoiding unnecessary verification of clearly identified goods.
Solution Approach 2:
The verification effort is distributed non-uniformly across different recognition results based on their extracted verification necessity. High-priority candidates that meet extraction conditions receive verification attention, while low-priority candidates are processed automatically. This local quality approach optimizes the balance between reliability and productivity by applying verification resources where they are most needed.
3Ease of operation
If image recognition system compares with registered images, then ease of operation is improved, but measurement precision deteriorates
Solution Approach 1:
The system performs preliminary registration of goods images in advance, creating a database for quick comparison during recognition operations. This preliminary action simplifies the operational process by having pre-prepared reference data, while the subsequent extraction and verification steps compensate for any accuracy limitations by focusing on uncertain cases.
Data Source
AI summary
An information processing system includes: a displayed goods recognition processing unit configured to recognize information indicating goods, the goods appearing in image information obtained by imaging a display shelf; and a verification target extraction processing unit configured to extract a recognition result to be a verification target out of the goods displayed on the display shelf, in which the verification target extraction processing unit extracts the verification target by using a change in the information indicating the goods displayed on the display shelf that has been recognized from the image information.