Defect Classification Using Deep Learning and Optical Coherence Tomography
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional defect classification systems for display devices require extensive human intervention to differentiate between false and true defects, leading to prolonged inspection and classification times.
Innovation Solution
A defect classification method utilizing a deep machine learning model, combining multi-optical vision and optical coherence tomography, where a first image of the exterior is analyzed to determine defects and extract XY coordinates, and a second image of the interior is used to train the model for accurate defect and type classification, employing a convolutional neural network with layers such as convolutional, pooling, and fully connected layers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If human determination is used to differentiate between false and true defects, then accuracy of defect classification is improved, but inspection time is significantly increased
Solution Approach 1:
The system performs preliminary automated defect detection using multi-optical vision to identify candidate defects and extract XY coordinates before deeper analysis. This preliminary action filters out obvious false defects early, reducing the workload for subsequent deep learning analysis and maintaining high accuracy while reducing overall inspection time.
Solution Approach 2:
A deep machine learning model trained on optical coherence tomography images serves as an intermediary between automated defect detection and human judgment. The model analyzes internal structures to determine whether detected defects are true or false, replacing the need for human determination while maintaining high classification accuracy and significantly reducing inspection time.
2Productivity
If deep machine learning model is used for automated defect determination, then inspection time is reduced, but system complexity is increased
Solution Approach 1:
The inspection system is segmented into distinct functional modules: multi-optical vision device for initial defect detection, optical coherence tomography device for internal structure imaging, and deep machine learning model for classification. Each module performs a specific function, making the complex system manageable and maintainable while achieving high inspection speed.
Solution Approach 2:
The deep machine learning model serves multiple functions: it classifies defects as true or false, determines defect types, and provides confidence scores for automated decision-making. This multi-functionality reduces the need for separate systems for each task, managing overall system complexity while maintaining high productivity.
3Measurement precision
If optical coherence tomography is used to image internal structure, then defect detection accuracy is improved, but measurement time is increased
Solution Approach 1:
Instead of performing full optical coherence tomography scanning of the entire display device, the system applies the technique locally to specific regions of interest identified by the multi-optical vision device. This localized approach maintains high defect detection accuracy by focusing imaging resources only where defects are suspected, while significantly reducing overall imaging time.
Solution Approach 2:
The multi-optical vision device performs preliminary defect detection to identify candidate defect locations and extract XY coordinates before optical coherence tomography imaging is performed. This preliminary action guides the subsequent high-precision imaging to only necessary regions, maintaining accurate defect detection while minimizing imaging time through targeted analysis.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach significantly reduces the time and increases the accuracy of defect inspection and classification by automating the process with deep machine learning, allowing for rapid and precise identification of defects within the display device.
Implementation Method 1
collecting a second image of an inside of the display device based on the XY coordinates of the defect of the display device by an optical coherence tomography device
Data Source
AI summary
A defect classification method includes collecting a first image of an exterior of a display device; determining a defect of the display device based on the first image; extracting XY coordinates of the defect of the display device; collecting a second image of an inside of the display device based on the XY coordinates of the defect of the display device; training a deep machine learning model for determining the defect of the display device and a defect type of the display device based on the second image; determining the defect of the display device based on the second image through the deep machine learning model; and determining the defect type of the display device based on the second image through the deep machine learning model.