Defect Image Classification Using Feature Clustering for Overlapping Faults
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing visual inspection methods for defect detection in semiconductor manufacturing vary in accuracy among individuals and are time-consuming when dealing with large volumes of image data, and existing neural network-based fault identification techniques suffer from reduced accuracy when multiple faults overlap.
Innovation Solution
A classification device and method utilizing a memory unit, processing unit, and classifier, employing a discriminative model with convolutional and fully connected layers to extract feature values from image data, followed by clustering to improve defect identification and classification accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If visual inspection is performed by individuals, then defect detection can be conducted, but accuracy varies among individuals and time consumption increases with large volumes of image data
Solution Approach 1:
The patent replaces manual visual inspection with an automated inspection device that captures images and uses image processing algorithms to detect defects. This substitution eliminates human variability and significantly reduces inspection time while maintaining or improving detection accuracy.
Solution Approach 2:
The patent introduces an image processing device as an intermediary between the captured image and the final defect identification. This intermediary extracts feature values from the image data and performs clustering analysis to objectively classify defects, removing the need for human judgment in the detection process.
2Productivity
If a neural network is used for fault identification, then automated defect detection is achieved, but accuracy decreases when multiple faults overlap
Solution Approach 1:
The patent segments the fault identification process into multiple stages: first extracting multiple types of feature values from different aspects of the image data, then performing clustering analysis to classify defects. This segmentation allows the system to handle overlapping faults by analyzing multiple characteristics simultaneously rather than relying on a single neural network pass.
Solution Approach 2:
The patent transitions from traditional single-dimensional neural network classification to multi-dimensional feature space analysis. By extracting multiple feature values (e.g., shape, texture, position) and performing clustering in this expanded feature space, the system can distinguish overlapping defects that would be indistinguishable in a single-dimensional approach.
3Measurement precision
If information on faults is obtained using traditional image processing, then fault identification can be performed, but accuracy is lowered when multiple faults overlap in the same region
Solution Approach 1:
The patent segments the complex image processing into distinct modules: feature extraction, clustering analysis, and classification. This modular approach manages complexity by breaking down the processing pipeline into manageable stages, each handling a specific aspect of defect analysis.
Solution Approach 2:
The patent extracts multiple feature values beyond what a single traditional method would provide, gathering comprehensive information from different aspects of the image data. This excessive extraction of features ensures that even when faults overlap, sufficient information is available to distinguish and classify each defect type accurately.
Data Source
AI summary
A novel classification device is provided. The classification device includes a memory unit, a processing unit, and a classifier. A plurality of pieces of image data and a discriminative model are stored in the memory unit. Each of the plurality of pieces of image data is image data determined to contain a defect. The discriminative model includes an input layer, an intermediate layer, and an output layer. First to n-th (n is an integer greater than or equal to 2) image data of the plurality of pieces of image data are supplied to the processing unit. The processing unit has a function of outputting feature values of the first to the n-th image data (a first to an n-th feature value) on the basis of the discriminative model. A feature value output from the processing unit is a numerical value of a neuron included in the intermediate layer. The first to the n-th feature value output from the processing unit are supplied to the classifier. The classifier has a function of performing clustering of the first to the n-th image data on the basis of the first to the n-th feature value.