Defect Image Generation Using Pattern Superposition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In manufacturing environments, particularly in semiconductor production, it is challenging to acquire defect images due to low defect incidence, leading to difficulties in generating high-quality defect images using existing methods like computer graphics or simulation-based approaches.
Innovation Solution
A device and method for generating defect images using a processor that stores normal and defect patterns, generates layout images, and integrates normal and defect patterns to create composite and defect images, respectively, facilitating the generation of various types of defect images with minimal data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If computer graphics or simulation-based methods are used to generate defect images, then defect images can be acquired for training AI models, but significant time and resources are required for high-quality 3D modeling and actual environments cannot be completely imitated
Solution Approach 1:
The patent uses real defect images captured from actual manufacturing environments as templates, copying their visual characteristics and defect patterns directly rather than creating synthetic 3D models. This allows defect images to be generated by superimposing these real defect patterns onto normal product images, significantly reducing the time and resources required compared to traditional 3D modeling approaches while maintaining authenticity of defect appearances
Solution Approach 2:
The patent pre-extracts and stores defect patterns from real defect images in advance, creating a library of reusable defect templates. When defect images are needed for AI training, these pre-prepared patterns can be quickly superimposed onto normal images without requiring time-consuming 3D modeling processes, thus resolving the contradiction between data quantity and time investment
2Quantity of substance
If computer graphics or simulation-based methods are used to generate defect images, then defect images can be acquired for training AI models, but significant resources are required for high-quality 3D modeling
Solution Approach 1:
Instead of investing significant computational resources in creating high-quality 3D models of defects, the patent copies actual defect patterns directly from captured defect images. This approach uses minimal computational resources while generating authentic-looking defect images suitable for AI model training, effectively resolving the resource consumption issue
Solution Approach 2:
The patent uses simple 2D image superposition techniques instead of computationally expensive 3D modeling. The defect patterns are treated as reusable templates that can be quickly applied to multiple normal images, providing an economical and efficient solution that requires minimal computational resources compared to traditional simulation-based methods
3Reliability
If real defect images are acquired from industrial environments, then authentic defect data can be obtained for AI training, but it is difficult to acquire due to security concerns and low defect incidence
Solution Approach 1:
The patent uses a small number of captured defect images as templates to generate a large quantity of defect images through pattern superposition. The system serves itself by using the limited authentic defect data it has acquired to create additional training data, thus resolving the contradiction between data authenticity and data quantity without requiring continuous access to rare real defect occurrences
Solution Approach 2:
The patent performs preliminary extraction and storage of defect patterns from authentic defect images in advance. These extracted patterns retain the authenticity of real defects and can be reused multiple times to generate additional training images, allowing the system to maintain data reliability while overcoming the limitation of low defect incidence in industrial environments
Data Source
AI summary
Provided are a device and method for generating a defect image. The device includes a memory configured to store a plurality of normal patterns with different sizes and a plurality of defect patterns and a processor configured to generate a layout image including at least one layout, generate a composite image by acquiring at least one normal pattern corresponding to a physical feature of each layout of the layout image from the memory, and generate a defect image by integrating at least one defect pattern with the composite image.