Automated SEM Sampling for E-Beam Image Enhancement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current imaging technologies, such as scanning electron microscopes (SEMs), face challenges in achieving high image quality while maintaining high throughput, especially in inspecting small defects on integrated circuit components.
Innovation Solution
The development of a method and apparatus for automatically obtaining training images to train a machine learning model, which analyzes patterns of data to identify training locations on a product sample, obtains both low and high quality images at these locations, and uses these images to enhance image quality without requiring extensive human intervention.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high quality images are obtained through multiple scans, then image quality is improved, but inspection throughput is reduced
Solution Approach 1:
The system performs preliminary actions by automatically selecting training locations based on product layout patterns before inspection begins. It pre-determines which locations need high-quality images for training the machine learning model, allowing the inspection process to be optimized in advance and reducing the need for multiple scans across the entire product surface.
Solution Approach 2:
The patent applies local quality by obtaining high quality images only at specific training locations identified through pattern analysis, rather than uniformly across the entire product. This selective approach ensures high image quality where needed for model training while maintaining higher throughput at other locations.
Solution Approach 3:
The system uses machine learning models to generate synthetic or enhanced images that replicate the appearance of high-quality physical images. The trained model can reproduce high-quality image characteristics from lower-quality input images, effectively copying the quality enhancement without requiring additional physical scanning.
2Adaptability or versatility
If manual image acquisition and processing is used, then flexibility is maintained, but operator variability and errors increase
Solution Approach 1:
The system performs self-service by automatically analyzing product layout patterns, selecting training locations, acquiring images, and training the machine learning model without requiring manual intervention. The automated workflow eliminates operator variability while maintaining adaptability through the machine learning model's ability to handle different product types and patterns.
Solution Approach 2:
The system incorporates feedback loops where the machine learning model continuously improves by analyzing the relationship between input images and ground truth labels. The model learns from the discrepancies between automated and manual assessments, progressively improving inspection accuracy and reducing errors over time.
Data Source
AI summary
Disclosed herein is a method of automatically obtaining training images to train a machine learning model that improves image quality. The method may comprise analyzing a plurality of patterns of data relating to a layout of a product to identify a plurality of training locations on a sample of the product to use in relation to training the machine learning model. The method may comprise obtaining a first image having a first quality for each of the plurality of training locations, and obtaining a second image having a second quality for each of the plurality of training locations, the second quality being higher than the first quality. The method may comprise using the first image and the second image to train the machine learning model.