Dr. Farshid Pirahansiah
Cookies support analytics and personalized ads. Accept, or only necessary storage is used.

By Dr. Farshid Pirahansiah — computer vision & embedded edge AI consultant, Berlin, Germany · 12+ years. Updated 2026-09-19.

Adaptive multi-threshold approach using MLP for license plate recognition systems.

Multi-threshold_Approach_for_License_Plate_Recognition_System

https://www.pirahansiah.com/notes/pubs/papers/Multi-threshold_Approach_for_License_Plate_Recognition_System

spotify

PDF Download My Conference Paper

My Conference Paper Multi-threshold Approach for License Plate Recognition System

My Conference Paper: Multi-threshold Approach for License Plate Recognition System

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.

Source: Pirahansiah, F. “Multi-Threshold for License Plate Recognition”. pirahansiah.com, updated 2026-09-19. Canonical URL https://pirahansiah.com/farshid/content/papers-multi-threshold-plate/ · markdown source papers-multi-threshold-plate.md · how to cite, quote and link this site.