Autonomous Reference Image Selection for Defect Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current defect detection methods are inefficient and resource-intensive, particularly in generating defect-free reference images and can produce ambiguous results when comparing evaluated objects to arbitrary reference objects.
Innovation Solution
A cluster-based and autonomous method for finding reference information that processes untagged images of items without prior tagging, using a repetitive clustering process with non-item specific neural networks to identify centroids, cluster features, and select reference images based on similarity, allowing for efficient defect detection without prior knowledge or historic bias.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a defect-free reference object image is generated for comparison, then defect detection accuracy is improved, but time and computational resources are consumed
Solution Approach 1:
The system automatically selects reference images from the inspected objects themselves without requiring external defect-free reference objects. The neural network clusters features from inspected objects and selects representative images as references, allowing the system to serve its own reference needs without external intervention or pre-prepared reference databases.
Solution Approach 2:
Instead of using actual defect-free reference objects, the system creates synthetic reference images by copying and selecting from inspected objects that appear to be defect-free based on their feature clusters. These copied images serve as references for comparison, eliminating the need for physical defect-free reference objects.
2Loss of time
If an arbitrary reference object is used for comparison, then computational resources are reduced, but detection results become ambiguous
Solution Approach 1:
The system uses neural network-based feature clustering to analyze and evaluate the inspected objects themselves, creating a feedback loop where the inspection results directly inform reference selection. This feedback mechanism ensures that references are chosen based on actual object characteristics rather than arbitrary selection, maintaining detection clarity while reducing computational burden.
Solution Approach 2:
The system changes the parameter of reference selection from arbitrary or pre-defined to dynamically selected based on feature cluster analysis. By adjusting the selection criteria to match the specific characteristics of inspected objects through neural network analysis, the system maintains detection accuracy without requiring computationally expensive defect-free reference generation.
3Productivity
If traditional defect detection methods are used, then defect detection can be performed, but the process is inefficient and resource-intensive
Solution Approach 1:
The system segments the defect detection process into distinct stages: feature extraction using neural networks, clustering of features, and selection of representative images as references. This segmentation allows each stage to be optimized independently, improving overall efficiency by avoiding unnecessary computational operations in stages that do not contribute to the final detection result.
Solution Approach 2:
The neural network-based feature extraction and clustering system serves multiple functions simultaneously: it analyzes object characteristics, identifies defect patterns, selects reference images, and prepares comparison data. This multi-functionality eliminates the need for separate specialized processes, significantly reducing computational resource consumption while maintaining detection effectiveness.
Data Source
AI summary
A method for cluster-based and autonomous finding of reference information, the method may include obtaining a group of untagged images, each untagged image captures an instance of an item; wherein at least some of the untagged images capture different instances of the item; obtaining multiple sets of item pixels from the untagged images of the group, each set originated from an untagged image of the group and comprises multiple item pixels; determining item features of the item for each set, based on the multiple item pixels of the set and; repeating, until reaching an end condition the steps of: (a) selecting some of the sets as centroids; (b) clustering the item features of the some of the sets to provide clusters, wherein the clustering is based, at least in part, on the centroids; and (c) removing members of a cluster that has less members than another cluster; and defining untagged images that are associated with a member of any remaining cluster as reference images or as reference image candidates.