Deep Learning Difference Filter and Aperture Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for selecting difference filters and aperture settings in image acquisition are manual, time-consuming, and prone to errors, particularly when dealing with multiple types of defects or multi-mode imaging, as they rely on iterative and intuition-based approaches.
Innovation Solution
A system and method utilizing a deep learning network classifier to generate difference filter and aperture setting recipes by applying and classifying various filters and settings on training images, extracting convolution layer filters, and performing mathematical analyses to optimize image acquisition for efficient defect review and classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If manual iterative methods are used to select aperture settings, then the selection process allows user intuition and flexibility, but the process becomes time-consuming and tedious
Solution Approach 1:
The system performs preliminary classification of defects and pre-selection of optimal aperture settings before actual imaging. The controller classifies defects in training images and determines optimal aperture settings in advance, creating a lookup table that guides subsequent imaging operations without requiring real-time iterative testing
Solution Approach 2:
The system enables self-service automation where the controller automatically selects aperture settings based on defect classification without requiring manual user input. The system uses machine learning models to autonomously determine optimal settings, reducing both time consumption and reliance on user expertise
2Ease of manufacture
If pre-selected difference filter recipes are used, then the method provides a set of predefined options for users, but the approach remains manual and intuition-based leading to errors
Solution Approach 1:
The system replaces manual mechanical selection of difference filters with an automated computational system. The controller uses machine learning models to automatically select optimal difference filters based on defect classification, substituting human intuition with algorithmic decision-making to improve reliability while maintaining the availability of predefined filter options
3Extent of automation
If algorithm selector software is used to estimate difference filters, then the method provides automated estimation, but the approach is still prone to error and not optimized for multiple defect types
Solution Approach 1:
The system segments the filter selection process by creating separate optimization paths for different defect types. The controller classifies defects into categories and selects difference filters specifically optimized for each defect type, rather than using a single generic estimation approach. This segmentation enables the system to handle multiple defect types effectively while maintaining high automation
Solution Approach 2:
The system creates a universal difference filter selection mechanism that can handle multiple defect types through a single integrated controller. The machine learning model is trained on diverse defect types and can automatically select appropriate filters for any defect category, making the system versatile across different inspection scenarios
4Measurement precision
If iterative testing of all aperture setting combinations is performed, then the method ensures comprehensive evaluation of options, but the process becomes excessively time-consuming
Solution Approach 1:
The system performs preliminary classification of defects and pre-determination of optimal aperture settings before actual inspection. By classifying defects in training images and determining optimal settings in advance, the system creates a lookup table that enables rapid selection during production inspection, maintaining evaluation quality while dramatically improving throughput
Data Source
AI summary
A system for defect review and classification is disclosed. The system may include a controller, wherein the controller may be configured to receive one or more training images of a specimen. The one or more training images including a plurality of training defects. The controller may be further configured to apply a plurality of difference filters to the one or more training images, and receive a signal indicative of a classification of a difference filter effectiveness metric for at least a portion of the plurality of difference filters. The controller may be further configured to generate a deep learning network classifier based on the received classification and the attributes of the plurality of training defects. The controller may be further configured to extract convolution layer filters of the deep learning network classifier, and generate one or more difference filter recipes based on the extracted convolution layer filters.