National Institute of Technology Karnataka Surathkal

Department of Information Technology

IT 290 Seminar
Vehicle Number Plate Detection through live stream using Optical
Character Recognition (OCR)

Charan Gowda Chowgondanahalli Manjunatha

231IT017
B.Tech [IT]
Dept of Information Technology, NITK Surathkal
Email: charan.231it017@nitk.edu.in
 Index

 Introduction
 Literature Survey
 Methodology
 Results Analysis
 Conclusion and Future work
 References
 1. Introduction
 Automatic Vehicle Plate Recognition (AVPR):
◦ Technology designed to detect, capture and interpret vehicle license plates
◦ Uses cameras and image processing license plate information without
human intervention.

 Real-world Applications:
◦ Widely used in parking management, law enforcement and urban
plannings.
◦ Helps monitor vehicle movement, track offenders.

 Critical Need:
◦ A road accident happens every four minutes, highlighting the urgent need
for better traffic monitoring and control.
◦ AVPR can play a crucial role in reducing accidents by improving
enforcement and response times.
Core Technology:
• AVPR relies on Optical Character Recognition (OCR) to extract text from
vehicle license plate images.
• OCR technology enables the system to accurately read and convert plate
numbers into digital data.

Real-time Processing:
• Designed for real-time operation, providing immediate responses such as
issuing fines or alerting authorities.
• This quick processing helps maintain better traffic discipline and faster
enforcement.

Extended Applications:
• Used in analyzing traffic patterns and aiding urban planning.
• These applications help optimize resource usage and improve overall city
infrastructure.
 2. Literature Survey
It covers real-world AVPR implementations, advancements
in image processing, and optimization techniques for Optical
Character Recognition (OCR).
Author Methodology Observation

X. Zhai et al.[1] OCR with neural networks; 97.3% accuracy; meets real-
training with 3500 UK binary time recognition
images requirements
Real-time mechanical Effective detection despite
P. Kulkarni et al.[2] identification for Indian inconsistent license plate
license plates standards

M.Y. Aalsalem et al.[3] AVPR for parking Efficient management and

management using monitoring of parking
surveillance cameras facilities
OCR Integration: Neural networks and motion analysis are integrated with traditional OCR
techniques to enhance accuracy and speed.
 For example, X. Zhai et al. achieved 97.3% accuracy in controlled environments by
combining these methods.

Context-Aware Adaptation: Tailoring systems like Har cascade classifier to handle non-
standard license plate formats common in countries like India is critical.
P. Kulkarni et al. demonstrated that custom adaptations are necessary for effective
operation under varied conditions.

Urban Applications: Beyond simple detection, AVPR can be integrated into parking
management systems to alleviate congestion.
M. Y. Aalsalem et al. showcased the potential benefits of such integration for efficient
parking and traffic flow.
 3. Methodology
The goal in this project is to automate the recognition of vehicle
registration numbers by reading number plates from live video feeds. To
achieve this, the project utilizes the open-source Python library Pytesseract,
which serves as a wrapper for the Google Tesseract OCR engine

FIG. Methodology followed in the project

1. Frame Acquisition:
• Live video capture with a camera placed 4–5 meters from the vehicle
• Captured in RGB format to preserve key color information
• Motion detection and frame filtering to select only frames with vehicles

2. Pre-Processing:
• Convert images to grayscale to reduce complexity
• Use Otsu’s thresholding and morphological operations (dilation/erosion)
• Apply binarization to create high-contrast images, isolating the number plate
from the background

3. Plate Localization:
• Identify candidate regions using a pre-defined aspect ratio
• Extract the isolated number plate area for further analysis
4. Character Segmentation:
•Segment the extracted plate image into individual characters
•Validate each segment based on area, shape, and texture
•Employ adaptive segmentation to handle overlapping characters

5. Character Recognition:
•Use Pytesseract to convert segmented characters into text
•Apply post-processing (pattern matching, feature extraction) for accuracy
•Display the recognized registration number on the terminal and integrate with
other systems
 4. Results and Analysis

System effectively identifies and extracts vehicle license plates in

real-time
Accuracy: 94.7%
Loss: 0.0316
F1 Score: 0.9911

Formula Used:
• Accuracy = (TN + TP) / (TP + FP + TN + FN)
• Precision = TP / (TP + FP)
• Recall = TP / (TP + FN)
• F1 Score = Harmonic mean of precision and
recall
 5. Conclusion and Future Work

 Achievements:
◦ Implementation of real-time AVPR system with high accuracy
◦ Effective multi-stage processing pipeline
◦ Practical application for traffic management and law enforcement

 Future Improvements:
◦ Dataset Expansion: Include more diverse real-world data
◦ Advanced Neural Networks: Incorporate CNNs and
transformer models
◦ Computational Efficiency: Optimize for edge devices
◦ Extended Applications: Integrate with urban management
systems
 6. References
[1] X. Zhai, F. Bensaali and R. Sotudeh, "OCR-based neural network for ANPR," 2012
IEEE International Conference on Imaging Systems and Techniques Proceedings,
Manchester, UK, 2012, pp. 393-397, doi: 10.1109/IST.2012.6295581.

[2] P. Kulkarni, A. Khatri, P. Banga and K. Shah, "Automatic Number Plate Recognition
(ANPR) system for Indian conditions," 2009 19th International Conference
Radioelektronika, Bratislava, Slovakia, 2009, pp. 111-114, doi:
10.1109/RADIOELEK.2009.5158763.

[3] M. Y. Aalsalem, W. Z. Khan and K. M. Dhabbah, "An automated vehicle parking

monitoring and management system using ANPR cameras," 2015 17th International
Conference on Advanced Communication Technology (ICACT), PyeongChang, Korea
(South), 2015, pp. 706-710, doi: 10.1109/ICACT.2015.7224887.
 [4] M. Y. Aalsalem and W. Z. Khan, "CampusSense — A smart vehicle parking monitoring
and management system using ANPR cameras and android phones," 2017 19th
International Conference on Advanced Communication Technology (ICACT),
PyeongChang, Korea (South), 2017, pp. 809-815, doi: 10.23919/ICACT.2017.7890230.

[5] M. G. Khan, Salma, M. Saeed, A. Zulfiqar, Y. Y. Ghadi and M. Adnan, "A Novel Deep
Learning Based ANPR Pipeline for Vehicle Access Control," in IEEE Access, vol. 10, pp.
64172- 64184, 2022, doi: 10.1109/ACCESS.2022.3183101.

[6] S. Shrivastava, S. K. Singh, K. Shrivastava and V. Sharma, "CNN based Automated

Vehicle Registration Number Plate Recognition System," 2020 2nd International
Conference on Advances in Computing, Communication Control and Networking
(ICACCCN), Greater Noida, India, 2020, pp. 795- 802, doi:
10.1109/ICACCCN51052.2020.9362737.
[7] C. J. Fernandez and R. R. Porle, "Raspberry Pi based ANPR for Smart
Access," 2021 IEEE International Conference on Artificial Intelligence in
Engineering and Technology (IICAIET), Kota Kinabalu, Malaysia, 2021, pp. 1-5,
doi: 10.1109/IICAIET51634.2021.9573652.

[8] Ball, K., 2018. The cases: ANPR, credit scoring and Neighborhood Watch 1.
In Surveillance and Democracy in Europe (pp. 30-50). Routledge.

[9] Sinha, H., Soumya, G.V., Undavalli, S., Jeyanthi, R. (2021). An Effective Real-
Time Approach to Automatic Number Plate Recognition (ANPR) Using YOLOv3
and OCR. In: Paprzycki, M., Thampi, S.M., Mitra, S., Trajkovic, L., ElAlfy, ES.M.
(eds) Intelligent Systems, Technologies and Applications. Advances in
Intelligent Systems and Computing, vol 1353. Springer, Singapore.
https://doi.org/10.1007/978-981- 16-0730-1_20