ANPR - Automatic Number Plate Recognition

This project is an example of how we can extract the number plate of a car image, by applying a series of image processing techniques. Considering:

• The image contains just 1 car
• The plate pattern is Mercosul

Algorithm

The following techniques sequence was applied in order to achiev the expected result:

1. Convert to Grayscale;
2. Black Hat Morphologial Transformations - to reveal characters against light backgrounds;
3. Close operation to fill gaps and little areas - to identify bigger structures;
4. Calculate the magnitude gradient (of axis x) of the Black Hat Transformation result image; . Put the result into [0, 255] scale
5. Smooth the image, apply another close transformation and another binary threshold using Otsu's method.

How to Develop

1. Clone this repo
2. Go to project folder
3. Start coding

# clone this repo
git clone https://github.com/virb30/anpr.git anpr
# cd to project folder
cd anpr

How to Deploy

1. Login to Heroku Account
2. Create a new Heroku Project
3. Commit and push changes to heroku

# login to heroku account
heroku login
# create heroku project
heroku create [project-name]
# commit & push changes to heroku
git add .
git commit -m "deploy"
git push heroku master

Demo

https://number-plate-extractor.herokuapp.com/

Next Steps

Our goal is to continue to develop this project by adding some features in the future:

• Integrate with Google Vision AI (for testing purposes)