Adaptive Defect Detection via Image Patch Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current defect detection processes in manufacturing are inaccurate due to the failure to distinguish between different populations of manufactured items, leading to inconsistent quality assessment.
Innovation Solution
An adaptive method for defect detection that segments images into patches and sub-patches, determining similarity thresholds based on similarities with training images to accurately identify defects in manufactured items, using a computerized system to execute the method and store instructions for executing the method.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single defect detection process is used for all manufactured items, then the process is simple to implement, but the detection accuracy is poor due to inability to distinguish between different populations
Solution Approach 1:
The patent segments the manufactured items into different populations based on their unique population distributions. Each population receives a tailored defect detection process with population-specific parameters, rather than applying a single uniform detection process to all items. This segmentation enables accurate distinction between different item populations while maintaining manageable process complexity through automated classification.
Solution Approach 2:
The patent implements local quality by applying different detection parameters and criteria specific to each population of manufactured items. Instead of a uniform detection approach, the system adapts the detection process to match the unique characteristics of each population, thereby improving detection accuracy without requiring complete redesign of the entire detection system.
2Measurement precision
If population-specific defect detection processes are implemented for each manufactured item population, then the detection accuracy is improved, but the complexity of the detection process increases
Solution Approach 1:
The patent performs preliminary classification of manufactured items into their respective populations before the actual defect detection process. By pre-organizing items according to their population characteristics and storing population distribution data, the system eliminates the need for complex real-time decision-making during detection, thereby maintaining high accuracy while controlling overall process complexity.
Solution Approach 2:
The patent creates and stores reference models of population distributions for each manufactured item population. These copied population characteristics serve as templates that guide the defect detection process, allowing the system to apply population-specific detection criteria without requiring complex real-time analysis of each item's characteristics during detection.
3Measurement precision
If comprehensive similarity calculations are performed across all training images, then the most accurate defect identification is achieved, but the calculation complexity and processing time increase significantly
Solution Approach 1:
The patent segments the similarity calculation process by comparing test images only against training images from the same population category. Instead of performing exhaustive comparisons across all training images, the system divides the search space into population-specific subsets, dramatically reducing calculation complexity while maintaining accurate defect identification within each population context.
Solution Approach 2:
The patent applies local quality by performing similarity calculations using only the relevant subset of training images that match the population characteristics of the test image. This localized approach focuses computational resources on the most relevant comparisons, reducing overall processing time while preserving the accuracy needed for population-specific defect detection.
4Measurement precision
If image segmentation into patches and sub-patches is performed, then the context information is preserved and defect detection accuracy is improved, but the computational complexity increases
Solution Approach 1:
The patent segments images into patches and sub-patches to preserve local context information and population characteristics. This segmentation enables the system to analyze different regions of an image with population-specific criteria, improving defect detection accuracy by maintaining contextual relationships while breaking down complex images into manageable analysis units.
Data Source
AI summary
A method for adaptive concept generation based on image context, the method include generating concepts that includes similarity thresholds for evaluating the similarity of received image sub-patches to reference image sub-patches of known properties.