Overview
An automated defect inspection system for flat panel displays (FPD) using advanced image processing techniques to detect and classify various types of manufacturing defects.
Features
- Defect Detection: Identify scratches, spots, lines, and other anomalies
- Classification: Categorize defects by type and severity
- Template Matching: Reference-based comparison
- Feature Extraction: Advanced pattern recognition
- Statistical Analysis: Defect distribution reporting
Tech Stack
- Platform: MATLAB
- Image Processing: MATLAB Image Processing Toolbox
- Computer Vision: Template matching algorithms
- Data Analysis: Statistical analysis tools
Detection Methods
Template Matching
- Reference image comparison
- Normalized cross-correlation
- Multi-scale matching
- Rotation-invariant detection
Feature Extraction
- Edge detection (Canny, Sobel)
- Morphological operations
- Texture analysis
- Contrast enhancement
Classification Algorithms
- Threshold-based classification
- Feature-based categorization
- Size and shape analysis
- Defect severity grading
Defect Types Detected
- Scratches: Linear surface damages
- Spots: Point defects and contamination
- Lines: Mura lines and streaks
- Stains: Irregular contamination patterns
- Bubbles: Air bubble defects
Technical Approach
Preprocessing
- Image alignment and registration
- Noise reduction
- Contrast adjustment
- Region of interest (ROI) extraction
Detection Pipeline
- Load reference and test images
- Preprocessing and enhancement
- Template matching / feature extraction
- Defect candidate identification
- Classification and grading
- Report generation
Results
- Detection accuracy: >92%
- Processing time: <2 seconds per panel
- False positive rate: <5%
- Suitable for production line integration
Industrial Application
- Implemented in LCD/OLED panel manufacturing
- Reduced manual inspection time by 70%
- Improved quality control consistency
- Enabled early defect detection in production