OCR based Automatic Number Plate Recognition (ANPR) using Artificial Neural Network
The project implements an OCR based license plate recognition using artificial neural network. The project has three modules - Number Plate Localization, Character Segmentation and Character Recognition.
The Project is implemented in Python using Artificial Neural Network (ANN) for classification and uses libraries like Numpy, OpenCV, Pandas, Scikit-learn and Scikit-image for image processing tasks like plotting intensity histogram for number plate localization in input image and character segmentation from localized plate.
Results
| Accuracy (in %) | |
|---|---|
| Plate Localization | 78.9% |
| Character segmentation | 60% |
| Character recognition from segmented characters | 55.66% |
For detailed information see - poster.pdf and projectReport.pdf
Setup Steps
This Environment need to be setup by using "Ubuntu 16.04 LTS Base" image.
sudo apt-get install python2.7
sudo apt-get install python-pip
sudo python -m pip install numpy
sudo python2 -m pip install opencv-python
sudo python -m pip install scipy
sudo python -m pip install pandas
sudo python -m pip install tqdm
sudo python -m pip install hickle
sudo python -m pip install matplotlib
sudo python -m pip install imutils
sudo python -m pip install scikit-image
Before running the application download the training data file from the given URL - (https://drive.google.com/file/d/1dSHrk0BsC06fNIgI2PhMf02jyhgDqVlf/view?usp=sharing) and copy it to project folder, this file should be named as "full_train.csv". For implementation of the project, we chose the data set from - http://www.ee.surrey.ac.uk/CVSSP/demos/chars74k/EnglishFnt.tgz
To run the Application Use the following way to run it from cmd prompt: python anpr.py
Example: python anpr.py car1.jpg