Multi-threshold_Approach_for_License_Plate_Recognition_System

https://www.pirahansiah.com/farshid/portfolio/publications/Papers/Multi-threshold_Approach_for_License_Plate_Recognition_System

spotify

PDF Download My Conference Paper

Multi-threshold Approach for License Plate Recognition System

1. Introduction

• Objective: Propose an adaptive multi-threshold approach for image segmentation, specifically in object detection.
• Application: Malaysian License Plate Recognition (LPR) system.
• Challenge: Different types of license plates require varied detection techniques depending on the country.

2. Adaptive Multi-threshold Approach

• Key Method:
  • Multi Layer Perceptron (MLP) trained by backpropagation to optimize threshold values.
  • Finds optimum threshold values by analyzing the peak value from a graph of object count versus threshold ranges.
• Advantages:
  • Adaptive to different types of license plates, including single-line and double-line plates with varying fonts.

3. Comparison with Other Threshold Methods

• Other Techniques:
  • Kittler and Illingworth’s Threshold
  • Potential Difference
  • Otsu’s Method
• Performance:
  • The adaptive multi-threshold approach improves overall performance compared to these existing methods.

4. License Plate Recognition Process

• Steps:
  1. Image Segmentation: Applies multi-thresholding to separate the license plate and individual characters.
  2. Character Segmentation: Uses the threshold values to segment characters within the license plate region.
  3. Recognition: Classifies the segmented characters to extract the license plate number.

5. Conclusion and Future Work

• Key Findings:
  • The proposed adaptive multi-threshold method enhances the performance of the LPR system.
• Future Work:
  • Further improvements are underway to accommodate real-time system specifications.