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
The following techniques sequence was applied in order to achiev the expected result:
- Convert to Grayscale;
- Black Hat Morphologial Transformations - to reveal characters against light backgrounds;
- Close operation to fill gaps and little areas - to identify bigger structures;
- Calculate the magnitude gradient (of axis x) of the Black Hat Transformation result image; . Put the result into [0, 255] scale
- Smooth the image, apply another close transformation and another binary threshold using Otsu's method.
- Clone this repo
- Go to project folder
- Start coding
# clone this repo
git clone https://github.com/virb30/anpr.git anpr
# cd to project folder
cd anpr
- Login to Heroku Account
- Create a new Heroku Project
- 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
https://number-plate-extractor.herokuapp.com/
Our goal is to continue to develop this project by adding some features in the future:
- Integrate with Google Vision AI (for testing purposes)