Self-Calibrating Optical Inspection Without Golden Samples
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional optical inspection techniques are ineffective in custom or unique production processes where the design and fabrication of inspected units are constantly changing, requiring pre-set golden samples and time-consuming machine learning model training, which is costly and error-prone.
Innovation Solution
A computer-based system performs automated optical inspection by identifying and quantifying attributes of features in image data, comparing individual instances to a global average, and modifying the image to mark deviations exceeding a threshold, allowing self-contained inspection without golden samples or extensive dataset training.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional optical inspection techniques use golden sample comparison, then inspection accuracy is improved for mass-production, but the system becomes ineffective for custom production with constantly changing designs
Solution Approach 1:
The system performs self-calibration by automatically establishing a baseline from the first inspection image and subsequently comparing all images against this baseline, eliminating the need for external golden samples or manual model training for each production run
Solution Approach 2:
The inspection system dynamically adapts to changing production processes by allowing operators to quickly update the baseline through simple image capture rather than requiring complete retraining of machine learning models, enabling rapid response to design changes
2Reliability
If machine learning models are pre-trained on multiple labeled datasets, then inspection accuracy is improved, but the process becomes time-consuming and costly
Solution Approach 1:
The system performs preliminary calibration by capturing a baseline image during the first production run, which then serves as the reference for all subsequent inspections, eliminating the need for time-consuming model training before each production batch
Solution Approach 2:
Instead of training complex machine learning models, the system creates a simple baseline copy of the first inspection image and uses this as the reference standard for all future comparisons, dramatically reducing computational requirements and training time
3Reliability
If machine learning models are calibrated to address errors, then inspection accuracy is improved, but the process becomes error-prone and costly
Solution Approach 1:
The system automatically maintains calibration by comparing all inspection images against the established baseline, with operators able to update the baseline through simple image capture rather than requiring complex model recalibration procedures
Data Source
AI summary
A computer-based system may quantify, based on the plurality of instances of a feature indicated by image data, an attribute (e.g., a color, a shape, a material, a texture, etc.) of the plurality of instances of the feature. The system may also quantify an attribute of an instance of the feature of the plurality of instances of the feature. The system may modify the image data to indicate the instance of the feature if/when a value of the quantified attribute of the instance of the feature exceeds a value of the quantified attribute of the plurality of instances of the feature by a threshold. Functionality (e.g., defective, non-defective, potentially defective, etc.) of the unit may be classified based on the modified image data.