License_Plate_Recognition_with_Multi-Threshold_Based_on_Entropy

https://www.pirahansiah.com/farshid/portfolio/publications/Papers/License_Plate_Recognition_with_Multi-Threshold_Based_on_Entropy

spotify

PDF Download My Conference Paper

License Plate Recognition with Multi-threshold based on Entropy

1. Introduction

• Objective: Propose a multi-thresholding method for license plate recognition.
• Importance of Thresholding:
  • Simplifies image segmentation
  • Ensures robustness and accuracy in recognizing license plate characters.
• Challenges:
  • Selecting the correct threshold values for better segmentation results.

2. Entropy-based Thresholding

• Method:
  • Based on maximizing the cross entropy between the original image and the segmented image.
  • Entropy is treated as a probability distribution of the image histogram.
• Historical Background:
  • Originally proposed by Pun and later improved by Kapur for image segmentation.
  • Entropy-based thresholding is widely used for bi-level and multi-level thresholding.

3. Proposed Method

• Multi-thresholding Based on Maximum Entropy:
  • Selects several threshold values by maximizing the entropy.
  • Integrates partial ranges of the image histogram to achieve better segmentation.
• Comparison with Other Methods:
  • Compared to single-thresholding techniques based on maximum entropy.
  • Evaluates how multi-thresholding enhances the recognition accuracy.

4. License Plate Recognition Process

• Steps:
  1. Image Segmentation: Separates the license plate region using multi-thresholding.
  2. Character Segmentation: Applies the selected thresholds to segment individual characters.
  3. Recognition: Recognizes the segmented characters to produce the license plate number.

5. Results and Discussion

• Performance:
  • The proposed multi-threshold method outperforms single-thresholding techniques.
  • Shows improved segmentation results for varying lighting and environmental conditions.

6. Conclusion

• Key Findings:
  • Multi-thresholding based on entropy enhances the accuracy of license plate recognition.
• Implications:
  • Can be applied in real-world license plate recognition systems with better robustness.