Article Recognition Device Feature Point Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional article recognition devices fail to accurately specify the commodity area identified by identification information, leading to a suboptimal recognition rate in image recognition processes.
Innovation Solution
An article recognition device comprising a camera, memory, and processor that captures images of articles on a table, extracts feature points, identifies articles by reading identification information, and specifies article areas using stored dictionary information and feature points, with the processor calculating similarity and removing irrelevant feature points to recognize remaining articles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional image recognition processes are used without accurate article area specification, then the processing is simpler, but the recognition rate does not improve
Solution Approach 1:
The patent segments the image processing into distinct stages: first identifying articles with identification information and specifying their article areas, then using these specified areas to guide subsequent image recognition processing. This segmentation allows the system to focus computational resources on relevant areas, improving recognition rate while maintaining manageable processing complexity
Solution Approach 2:
The patent performs preliminary action by specifying article areas based on identification information before conducting image recognition. This preliminary specification of search areas enables the subsequent recognition process to focus only on relevant regions, thereby improving recognition accuracy without proportionally increasing overall processing complexity
2Measurement precision
If image recognition is performed on the entire captured image area, then the coverage is complete, but the recognition accuracy decreases due to irrelevant features
Solution Approach 1:
The patent extracts and removes feature points that fall within already-identified article areas from the set of feature points used for image recognition. This extraction of relevant features from the entire image area allows the system to maintain complete area coverage while improving recognition accuracy by eliminating redundant or irrelevant features from articles that have already been identified
Solution Approach 2:
The patent applies local quality by treating different areas of the image differently: areas containing identification information are processed to specify article boundaries, while other areas are processed for image recognition. This localized processing strategy ensures that each area is handled with the appropriate level of detail and focus, improving overall recognition accuracy
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
An article recognition device includes a table, a camera that captures an image of a predetermined area on the table, a memory that stores dictionary information indicating a predetermined set of feature points of each article, and a processor. The processor extracts feature points of the articles in the captured image, identifies a first article by reading identification information in the captured image, acquires the predetermined set of feature points of the identified article, specifies an article area of the first article in the captured image based on the extracted feature points and the predetermined set, removes feature points within the specified article area from the extracted feature points, and recognizes a second article based on the extracted feature points where the feature points within the specified article area have been removed and the predetermined set of feature points of the second article stored in the dictionary information.