License Plate Recognition

1
License Plate Recognition

• Artificial Intelligence Term Project

Abdoulaye Fadimatou Roger Blanco Ribera
2
Contents

• Introduction
• LPR system
• Plate Location
• Korean license plates
• Statistical Approach
• Edge Matching
• Plate segmentation
• Binarization
• Character Extraction
• Character Recognition
• Template Matching

3
LPR System

• LPR System
• Hardware
• Software
• Database
• Applications
• Law enforcement
• Access control
• Traffic monitoring

4
LPR System

• Our approach

5
Korean License Plates
6
Statistical Approach

• Use color information
• Illumination changes ? HSI model, Lab Model
• MLE estimation (Gaussian Assumption)
• Classify
• Threshold in Mahalanobis distance

7
Statistical Approach

• Train data
• 80 10x10 squares
• 3 different days
• Different lighting conditions

8
Edge Matching

• Use geometrical information
• Robust to illumination
• changes
• Prewitt Mask

9
Edge Matching

• Use geometrical information

• Possible candidate pairs
• use color information

• Special Ratio
• 21

10
Binarization

• K-means Algorithm
• RGB space Saturation

11
Extraction
12
Character Recognition

• Auto-correlation
• Cross-correlation with same size template
• Mismatch penalty

13
Results
14
Thank you for your attention

• Questions?
• Comments?