Project Report

License Plate Recognizer System

1. Project Overview
Title: License Plate Recognizer System
Team: [Muskan Pal (48)]

[Mangesh Lodhi (44)]

2. Abstract
The License Plate Recognizer (LPR) System aims to automatically detect and recognize vehicle license
plates using image processing techniques and Optical Character Recognition (OCR). This project
explores real-time recognition in various conditions (lighting, angles, motion). The system is built
using Python and MATLAB for tasks such as image preprocessing, license plate detection, character
segmentation, and recognition.

3. Scope
The License Plate Recognizer System has widespread applications in sectors such as:

 Traffic Law Enforcement: Automatically identifying vehicles involved in traffic violations.

 Parking Management: Automatic vehicle entry/exit based on pre-registered license plates.

 Toll Collection: Automating toll transactions.

 Access Control: Monitoring vehicles in restricted or high-security areas.

The scope of this project includes:

1. Building a prototype capable of recognizing license plates from static images and live video
streams.

2. Supporting recognition in challenging conditions (e.g., low light, different angles).

1. OpenALPR dataset: Contains images of vehicles with varied license plate types from different
countries.

2. SSIG dataset: Provides annotated vehicle images with varying lighting and environmental
conditions.

3. UFPR-ALPR dataset: A large dataset with over 4,000 vehicle images for automatic license
plate recognition.

Dataset Features:

 License plates from various regions with different plate formats.

 Annotated bounding boxes for license plate regions.

 High-resolution images to allow clear character segmentation.

The dataset will be split into:

 Training set: For training the model.

 Testing set: For evaluating the performance of the system in real-world conditions.

5. Methodology
The License Plate Recognizer System follows a structured methodology for image processing and
character recognition:

5.1 Image Preprocessing

1. Grayscale Conversion: Converting images to grayscale to reduce computational complexity.

2. Noise Reduction: Applying filters like Gaussian blur or median filtering to reduce noise.

3. Edge Detection: Using edge detection methods like Canny to identify the boundaries of the
license plate.

4. Thresholding: Binarizing the image using techniques like Otsu’s thresholding to distinguish
characters from the background.

5.2 License Plate Detection

1. Contour Detection: Locating potential rectangular regions in the image using contours.

2. Object Detection with Deep Learning: Optionally, using YOLO (You Only Look Once) or
similar models to detect the license plate region.

5.3 Character Segmentation

 Connected Component Analysis: Segmenting individual characters from the binarized plate
image.

5.4 Optical Character Recognition (OCR)

1. MATLAB's ocr() Function: Recognizes text from images of the license plate.

2. Tesseract OCR: Alternatively, Tesseract OCR in Python can be used for text recognition if
MATLAB is not available.

5.5 Post-processing

 Error Correction: Use error correction techniques (e.g., pattern matching, Levenshtein
distance) to fix misrecognized characters.

 String Validation: Validate recognized text against known license plate formats.

6. Tools and MATLAB Functions

1. MATLAB: Used for image processing, character segmentation, and OCR. MATLAB’s powerful
built-in functions make it suitable for developing a rapid prototype.

o Image Processing Toolbox: For preprocessing and plate detection.

o Computer Vision Toolbox: For object detection and text recognition.

2. Python (Optional): Can be used alongside MATLAB for tasks like deep learning (using
frameworks like (TensorFlow or PyTorch) and OCR (using Tesseract).

3. OpenCV: A Python library for advanced image processing, including preprocessing and plate
localization.

4. Tesseract OCR: Open-source tool used to recognize text on the license plate.

6.2 MATLAB Functions

1. imread(): Reads images into MATLAB for processing.

2. rgb2gray(): Converts color images to grayscale.

3. imadjust(): Adjusts image intensity for better contrast.

4. edge(): Performs edge detection using Canny or Sobel algorithms.

5. regionprops(): Measures properties of connected components (used for character

1. Real-time License Plate Detection: A system capable of detecting license plates in static
images and live video streams with high accuracy.

2. Character Recognition: The ability to extract and recognize characters from license plates
using OCR with an accuracy of 85-95%, depending on the dataset and conditions.

3. Dataset Performance: Performance evaluation on publicly available datasets, with a focus on

4. Error Correction: Implementation of error correction techniques to improve recognition

5. System Scalability: The system should be scalable for real-world applications such as traffic
monitoring or parking management.

8. Conclusion
The License Plate Recognizer System combines traditional image processing techniques and modern
machine learning-based methods to create a reliable, real-time system for detecting and recognizing
license plates. Using MATLAB for preprocessing and OCR, the system can process images efficiently
and achieve high accuracy. This project serves as a foundational model that can be further improved
with advanced deep learning techniques for enhanced accuracy and robustness under varying
environmental conditions.

9. Future Work
1. Deep Learning-Based OCR: Implement deep learning-based OCR models to improve
recognition accuracy in challenging conditions.

2. Mobile and Cloud Integration: Develop a mobile-based version of the system for real-time
monitoring, or integrate the system with cloud services for scalable processing.

3. Enhance Robustness: Improve performance in low-light and high-speed scenarios through

References
1. MATLAB Documentation: https://www.mathworks.com/help/matlab/

2. OpenCV Documentation: https://opencv.org/

3. Tesseract OCR: https://github.com/tesseract-ocr/tesseract

4. Public License Plate Datasets: UFPR-ALPR, OpenALPR, SSIG datasets.