Production Line Image Inspection With Selective Reinspection Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current automatic appearance inspection systems for electronic components face inefficiencies and inaccuracies due to high manpower consumption and potential missed inspections, especially when the number of defect training samples is small, leading to defective products being passed to clients.
Innovation Solution
A system for intelligently monitoring production lines that includes a training subsystem, an operating station host, and a classifier subsystem, which analyzes image features, updates classification decisions, and reduces reinspection through a labeling module and image correlation analysis, allowing for real-time updates and reduced manual verification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If automatic appearance inspection device is used, then inspection efficiency is improved, but inspection accuracy deteriorates due to wrong or missed inspection
Solution Approach 1:
The patent introduces an intermediary verification mechanism where operator reinspection serves as a mediator between the automatic inspection device and final product disposal. The operator reviews and corrects automatic inspection results, ensuring accuracy while maintaining automated efficiency. This intermediary layer resolves the contradiction by combining machine speed with human accuracy.
Solution Approach 2:
The system implements feedback loops where operator corrections and reinspection results are fed back to continuously improve the automatic inspection device's accuracy. The system learns from operator interventions and adjusts its classification thresholds, reducing missed inspections over time while maintaining high throughput.
2Reliability
If operator performs reinspection on all inspected object images, then inspection accuracy is improved, but manpower consumption increases
Solution Approach 1:
Instead of requiring operators to inspect all images, the system applies partial action by having operators review only specific cases: images with low confidence scores, borderline classifications, or unusual defect patterns. This selective reinspection approach maintains accuracy for critical cases while dramatically reducing overall manpower consumption.
Solution Approach 2:
The automatic inspection device performs self-service by initially processing all images and only flagging uncertain cases for human review. The system autonomously handles clear-cut cases without human intervention, serving itself for routine inspections and reserving operator resources for ambiguous cases that truly need human judgment.
3Device complexity
If number of defect training samples is small, then system complexity is reduced, but inspection accuracy deteriorates
Solution Approach 1:
The system performs preliminary action by collecting and curating training samples during the operator reinspection phase. As operators review and label images, these labeled samples are accumulated and used to progressively train and improve the automatic inspection model, enabling accurate detection even with initially small training datasets.
Solution Approach 2:
The system ensures continuity of useful action by continuously accumulating labeled defect samples from ongoing operations and continuously retraining the inspection model. This continuous learning process transforms the limited initial training data into an expanding knowledge base, progressively improving accuracy without requiring a complete retraining dataset.
Data Source
AI summary
System and method for intelligently monitoring the production line that can monitor an inspected object image captured by an image capturing device, thereby allowing an operating station host to provide a labeling module to reinspect a classification decision of a classifier subsystem, to achieve the purpose of verifying the classification decision or checking whether there are missed inspections. In addition, the classifier subsystem can automatically filter out classification decisions with lower reliability to effectively reduce the number of reinspection. Moreover, a group of inspected object images can be analyzed first to obtain image difference features through comparison, which is suitable for insufficient training samples. Furthermore, the labeling module can simultaneously reinspect highly relevant historical classification decisions. Meanwhile, a second image capturing device is provided, so that the system can automatically label defect positions based on the inspected object image before and after repair, thereby learning to judge whether defects occur.