AI Visual Inspection Data Quality Indexing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional AI models for visual inspection fail to accurately distinguish actual defects from false defects due to variations in data collection conditions such as lighting, camera alignment, and blur, leading to misleading performance and the identification of false defects.
Innovation Solution
A method and system that generates representative images to calculate differences in luminance values and determine indices representing data-capturing conditions, allowing for automatic correction of misalignment, illumination variations, and blur by adjusting image data during the AI-based visual inspection process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If AI models are trained on data collected under varying conditions (lighting, alignment, blur), then the system can handle diverse production scenarios, but the model misidentifies capturing condition variations as defects leading to false positives
Solution Approach 1:
The patent segments the image analysis process into two distinct stages: (1) capturing condition assessment using representative images and difference imaging to evaluate lighting, alignment, and focus conditions, and (2) defect detection. This segmentation allows the system to separately assess data quality before defect analysis, preventing misidentification of capturing condition variations as defects while maintaining adaptability to diverse production scenarios.
Solution Approach 2:
The patent introduces representative images and difference images as intermediary elements between the raw captured images and the defect detection process. These intermediary images serve as references to assess capturing conditions, acting as a mediator that separates the evaluation of data quality from defect identification, thereby reducing false positives while handling diverse production conditions.
2Productivity
If conventional AI models process images without checking capturing conditions, then the processing speed is fast and simple, but the models fail to distinguish actual defects from false defects caused by lighting and alignment variations
Solution Approach 1:
The patent implements preliminary assessment of capturing conditions using representative images and difference imaging techniques before the main defect detection process. By pre-evaluating lighting, alignment, and focus conditions, the system prepares quality metrics in advance, enabling fast and reliable defect detection without compromising processing speed. This preliminary action ensures that only images meeting quality thresholds proceed to defect analysis.
Solution Approach 2:
The patent replaces complex manual quality assessment mechanisms with automated image processing techniques. Instead of requiring manual inspection of capturing conditions, the system uses algorithmic comparison of representative images and difference imaging to automatically evaluate data quality, maintaining high processing speed while improving reliability through consistent, objective assessment criteria.
3Measurement precision
If the system implements comprehensive checking of data gathering conditions, then false defect identification is reduced, but the system complexity and computational requirements increase
Solution Approach 1:
The patent applies local quality assessment by focusing the representative image comparison and difference imaging analysis specifically on critical regions and key capturing condition parameters (lighting, alignment, focus) rather than进行全面 analysis of all image aspects. This localized approach maintains high defect detection accuracy while reducing overall system complexity by concentrating computational resources on the most impactful quality indicators.
Solution Approach 2:
The patent transforms the complex multi-dimensional problem of capturing condition assessment into simplified parameter comparisons by using difference imaging to quantify changes in lighting, alignment, and focus as specific measurable parameters. This parameterization approach enables precise defect detection through quantitative thresholds while keeping the system structure relatively simple and computationally efficient.
Data Source
AI summary
A method and system for checking data gathering conditions or image capturing conditions associated with images during AI based visual-inspection process. The method comprises generating a first representative (FR1) image for a first group of images and a second representative image (FR2) for a second group of images. A difference image data is generated between FR1 image and the FR2 image based on calculating difference between luminance values of pixels with same coordinate values. Thereafter, one or more of a plurality of white pixels or intensity-values are determined within the difference image based on acquiring difference image data formed of luminance difference-values of pixels. An index representing difference of data-capturing conditions across the FR1 image and the FR2 image is determined, said index having been determined at least based on the plurality of white pixels or intensity-values, for example, based on application of a plurality of AI or ML techniques.