Global Texture Feature Extraction for Semiconductor Defect Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current deep learning methods for classifying defects in semiconductor specimens focus on local patterns, neglecting global texture features that can improve classification performance.
Innovation Solution
A system and method that determine both global texture characteristics and local characteristics of a specimen's image, using these features as inputs to a machine learning model for enhanced classification and metrology applications.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep learning methods focus only on local patterns for defect classification, then the method complexity remains manageable, but classification accuracy deteriorates due to neglecting global texture features
Solution Approach 1:
The image processing is segmented into distinct feature extraction modules: local pattern extraction using convolutional neural networks and global texture characteristic extraction using separate texture analysis algorithms. This segmentation allows each module to specialize in specific features without overwhelming complexity, while the fusion of both feature types improves classification accuracy.
Solution Approach 2:
The patent merges local pattern features and global texture characteristics into a unified feature vector that serves as input to the classification model. This combination integrates complementary information from both local and global perspectives, resolving the contradiction by achieving high accuracy through comprehensive feature representation while managing complexity through modular architecture.
2Loss of information
If only local features are extracted from specimen images, then processing time and computational resources are reduced, but important global texture information is lost
Solution Approach 1:
Global texture characteristics are extracted in parallel with local feature extraction rather than sequentially, performing both operations simultaneously during the feature extraction phase. This preliminary action ensures no information is lost while avoiding sequential processing delays, as both feature types are computed concurrently and fed into the classification model together.
3Measurement precision
If high resolution imaging is used for defect review, then defect classification accuracy improves, but inspection throughput decreases
Solution Approach 1:
The system applies different quality levels of image processing to different regions and purposes: low-resolution images are processed using the new global texture feature extraction method for rapid classification, while high-resolution images are used only when detailed defect characterization is required. This local quality approach maintains high throughput for most inspections while preserving accuracy for critical cases.
Data Source
AI summary
Methods and systems for determining information for a specimen are provided. One system includes a computer subsystem configured for determining a global texture characteristic of an image of a specimen and one or more local characteristics of a localized area in the image. The system also includes one or more components executed by the computer subsystem. The component(s) include a machine learning model configured for determining information for the specimen based on the global texture characteristic and the one or more local characteristics. The computer subsystem is also configured for generating results including the determined information. The methods and systems may be used for metrology (in which the determined information includes one or more characteristics of a structure formed on the specimen) or inspection (in which the determined information includes a classification of a defect detected on the specimen).