Substrate Inspection Using ML Estimated Images to Filter Unevenness
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing defect inspection methods struggle with high accuracy due to unevenness in substrate images, which can lead to false defect detection.
Innovation Solution
A substrate inspection apparatus that uses an image estimation model created by machine learning to generate an estimated image of a substrate after processing, allowing for accurate defect detection by comparing the captured image with the estimated image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional defect inspection methods are used, then the inspection process is simple, but defect detection accuracy is low due to unevenness in substrate images causing false defect detection
Solution Approach 1:
The system performs preliminary actions by capturing images at multiple process stages (before and after processing) and using machine learning to generate estimated images that predict the appearance of substrates after processing. This preliminary estimation allows for comparison with actual captured images to identify true defects while filtering out false positives caused by processing-induced unevenness.
Solution Approach 2:
The machine learning model acts as an intermediary that processes captured images and generates estimated images serving as a reference. This intermediary component bridges the gap between raw captured images and defect identification, enabling accurate defect detection by comparing actual images against the estimated reference images that account for expected processing variations.
2Measurement precision
If the inspection threshold is lowered to improve defect detection sensitivity, then more defects are detected, but false positives from unevenness increase
Solution Approach 1:
The system implements feedback by comparing captured images against estimated images generated by the machine learning model. This comparison provides feedback that distinguishes between actual defects and normal processing variations, allowing the system to maintain low inspection thresholds for high sensitivity while filtering out false positives through the reference estimation feedback mechanism.
Solution Approach 2:
The machine learning model creates copies of substrate images at different process stages and generates estimated images that replicate the expected appearance after processing. These copied and estimated images serve as reference standards for comparison, enabling the system to identify true defects while ignoring normal variations that appear in the copying and estimation process.
Data Source
AI summary
A substrate inspection apparatus for inspecting a substrate, includes: an acquisition part configured to acquire an estimated image of an inspection target substrate after a process by a substrate processing apparatus, based on an image estimation model created by a machine learning by using a captured image before the process by the substrate processing apparatus and a captured image after the process by the substrate processing apparatus for each of a plurality of substrates, and a captured image of the inspection target substrate before the process by the substrate processing apparatus; and a determination part configured to determine the presence or absence of a defect in the inspection target substrate, based on a captured image of the inspection target substrate and the estimated image of the inspection target substrate after the process by the substrate processing apparatus.