Project 3: License Plate Recognition (ANPR)

Automatic Number Plate Recognition system for parking lots, toll booths, and security systems.

What You’ll Learn

  1. Plate Detection - Using Haar Cascades and contour analysis
  2. Image Preprocessing - Preparing plates for OCR
  3. OCR Integration - Extracting text with EasyOCR
  4. Format Validation - Verifying plate formats
  5. Logging System - Recording vehicle entries/exits

Key OpenCV Functions

Function Purpose
cv2.CascadeClassifier() Haar cascade detection
cv2.bilateralFilter() Noise reduction
cv2.Canny() Edge detection
cv2.findContours() Find plate boundaries
cv2.adaptiveThreshold() Prepare for OCR

Usage 

# Run demo with generated plate
python main.py --demo

# Process a single image
python main.py --image /path/to/car.jpg

# Use webcam for real-time detection
python main.py --camera

# Process video file
python main.py --video /path/to/traffic.mp4

Algorithm Flow 

1. Input Image/Video Frame
        |
2. Plate Detection
   ├── Haar Cascade
   └── Contour Analysis (fallback)
        |
3. Plate Preprocessing
   ├── Resize
   ├── Grayscale
   ├── Bilateral Filter
   └── Adaptive Threshold
        |
4. OCR (EasyOCR)
        |
5. Format Validation
        |
6. Logging & Display

Detection Methods

Haar Cascade

  • Fast and lightweight
  • Uses pre-trained cascade classifier
  • Works well for frontal plates

Contour Analysis

  • Finds rectangular contours
  • Checks aspect ratio (2.0 - 6.0)
  • Works for various plate orientations

Real-World Applications

  • Parking lot management
  • Toll collection systems
  • Traffic monitoring
  • Security checkpoints
  • Law enforcement
  • Access control gates

Code Highlights

Plate Detection 

# Contour-based detection
aspect_ratio = w / float(h)
if 2.0 < aspect_ratio < 6.0:  # Plate aspect ratio
    plate_candidates.append((x, y, w, h))

OCR Preprocessing 

# Adaptive thresholding for clean OCR input
thresh = cv2.adaptiveThreshold(
    filtered, 255,
    cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
    cv2.THRESH_BINARY_INV,
    19, 9
)

Output Format

Log File (CSV) 

PlateNumber,Confidence,Timestamp
ABC1234,0.92,2024-01-15 09:00:15
XYZ5678,0.87,2024-01-15 09:02:33

Performance Tips

  1. Use cascade detection first, contour as fallback
  2. Process every N frames in video
  3. Implement cooldown for same plate
  4. Use GPU-accelerated OCR if available

References