Wafer Image Screening for Accurate Defect Determination
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional wafer defect detection methods face challenges in accurately distinguishing defects due to abnormal images caused by photographing apparatus issues, leading to reduced accuracy in defect determination and product quality.
Innovation Solution
A determination method that excludes images deemed as misdetermination candidates, using a model to identify and exclude images with missing areas or unwanted features, and employs a composite determination model to accurately assess wafer acceptability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If captured images are used for defect determination, then automated defect detection can be performed, but images with abnormalities from the photographing apparatus may cause misdetermination and reduce accuracy
Solution Approach 1:
The patent applies preliminary action by performing quality assessment of captured images before they are used for defect determination. The image quality assessment unit evaluates each captured image to determine if it is suitable for accurate defect detection, excluding images with abnormalities such as missing areas or unwanted features. This preliminary filtering ensures that only high-quality images proceed to the defect determination stage, thereby maintaining high measurement precision while enabling automated detection.
2Productivity
If all captured images are included in defect determination, then processing speed is maintained, but accuracy decreases due to inclusion of misdetermination candidate images
Solution Approach 1:
The patent applies the taking out principle by extracting and removing misdetermination candidate images from the set of captured images before defect determination. The image quality assessment unit identifies images with abnormalities (such as missing areas, unwanted features, or poor quality characteristics) and excludes them from the defect determination process. This selective extraction ensures that only reliable images are used, improving accuracy without significantly impacting processing speed since the assessment is performed efficiently in the automated workflow.
3Productivity
If visual inspection is replaced by automated detection, then labor hours are reduced, but accuracy may decrease due to photographing apparatus abnormalities
Solution Approach 1:
The patent applies feedback by implementing an image quality assessment mechanism that provides information about the quality of each captured image to the defect determination process. The image quality assessment unit evaluates images and provides feedback on their suitability for defect detection, allowing the system to adjust its processing accordingly. This feedback loop ensures that automated detection maintains high accuracy by identifying and excluding images that would lead to misdetermination, while still achieving the labor efficiency benefits of automation.
Data Source
AI summary
A determination method for a wafer 30 includes: obtaining a captured image 40 of at least a portion of a wafer 30 as a determination image to be used to determine whether the wafer is acceptable or not, excluding the captured image 40 from the determination image when the captured image 40 corresponds to a misdetermination candidate image, and determining whether the wafer 30 is acceptable or not based on the determination image.