Hybrid Article Inspection Using AI and Rule-Based Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing article inspection apparatuses struggle with inaccurate detection of foreign substances and flaws due to limitations in distinguishing similar characteristics or unpredictable defects.
Innovation Solution
An article inspection apparatus that combines AI processing and rule-based image processing to enhance detection accuracy by using a learning model for AI processing and a rule-based processing unit, with a comprehensive determination unit to synthesize results and set determination rules based on article type and processing accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a learning model is used to detect defects, then detection accuracy for trained defects is improved, but detection reliability for unpredictable defects deteriorates
Solution Approach 1:
The patent combines AI processing (learning model) and rule-based processing into a hybrid inspection system. The AI processing unit handles trained defect patterns while the rule-based processing unit handles unpredictable defects using predetermined algorithms. Both processing results are integrated by the comprehensive determination unit to achieve both high accuracy for known defects and high reliability for unknown defects.
2Measurement precision
If image processing filters are applied to detect foreign substances, then detection capability is improved, but false detection of similar characteristics deteriorates
Solution Approach 1:
The patent introduces an intermediary comprehensive determination unit that mediates between AI processing results and rule-based processing results. This intermediary synthesizes both processing outcomes, comparing and validating findings to reduce false detections while maintaining high detection capability for foreign substances with similar characteristics to the article.
Data Source
AI summary
Provided is an article inspection apparatus that improves inspection accuracy. An AI processing unit outputs determination data indicating a quality of an article captured in a captured image using a learning model that has been trained with an image, and a rule-based processing unit outputs determination data indicating the quality of the article captured in the captured image using an image processing unit that performs image processing on the captured image. A comprehensive determination unit performs comprehensive determination of the quality of the article captured in the captured image based on both the determination data output from the AI processing unit and the rule-based processing unit.