Project 5: AI-Powered Quality Inspection System

Industrial defect detection system - detect cracks, scratches, or anomalies in products.

What You’ll Learn

  1. Image Preprocessing - CLAHE, Gaussian blur
  2. Defect Detection - Multiple methods
  3. Template Matching - Comparison with reference
  4. Pass/Fail Logic - Quality decision making
  5. Report Generation - Documenting results

Detection Methods

Method Best For
Threshold Dark spots, stains
Edge Detection Cracks, scratches
Blob Detection Holes, bubbles
Comparison Any deviation from reference

Usage 

# Run demo with sample images
python main.py --demo

# Use camera for real-time inspection
python main.py --camera

# Inspect specific image
python main.py --image product.jpg

# With reference image
python main.py --image test.jpg --reference good.jpg

Algorithm Flow 

1. Load test image
        |
2. Preprocess
   ├── Grayscale
   ├── CLAHE enhancement
   └── Gaussian blur
        |
3. Detect defects
   ├── Adaptive threshold
   ├── Canny edges
   ├── Blob detection
   └── Reference comparison
        |
4. Classify defects
        |
5. Make decision (pass/fail)

Key OpenCV Functions

Function Purpose
cv2.createCLAHE() Contrast enhancement
cv2.adaptiveThreshold() Local thresholding
cv2.Canny() Edge detection
cv2.SimpleBlobDetector() Blob detection
cv2.absdiff() Image comparison

Real-World Applications

  • Manufacturing quality control
  • PCB inspection
  • Food product inspection
  • Textile defect detection
  • Pharmaceutical packaging

Code Highlights

CLAHE Enhancement 

clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8, 8))
enhanced = clahe.apply(gray)

Reference Comparison 

diff = cv2.absdiff(reference_gray, test_gray)
_, binary = cv2.threshold(diff, threshold, 255, cv2.THRESH_BINARY)

Quality Thresholds

Parameter Default Description
max_defects 3 Max allowed defects
max_area 1000px Max total defect area
critical_types [‘crack’] Auto-fail defect types

References