Licence-Plate-Detection

The code is an approach to Detect licence plate of vehicles with use of Machine Learning Algorithms and Image Processing techniques

A sample of Dataset is provided: Look under the directory folder to get a sense of the dataset.

• You would want to start with the module LP_Detect_main.py. Let go one by one

  create_feature_matrix:

  • Calls a function of the class CrtFeatures: inside the module Bld_FeatureCrps. This module is aimed to extract features from a license plate/non license plate and store the features into disk (Training Features). It basically uses two directories
    • Train_data1 : Manually extracted license plates (cropped form vehicles image).
    • Train_data2 : Some random images (Non License plate) -- see config-local.conf file to get to know the directories name

  train_model:

  • Makes call to function of class Model inside module BldModel. Now that we have extracted training features as discussed above, we would want algorithm to learn the features of a license plate and a non-license plate. This is achieved by train_model.
    • It fetches the saved features and the corresponding label (license_plate or non-license plate) and sends it to the SVM model.
    • It also stores the learned model into the disk so that while cross validation and testing we can invoke the model and classify.

  valid and run_cross_valid:

  • These function makes use of the above modules to extract features of a manually cropped license plate/non-license plate and use the saved model to classify the images.

  Extract_lisenceplate:

  • This is the most important function that is provided with the actual directory where your images (vehicles/non-vehicles) are provided.

    • It extracts all contours (rectangles, circles polygon defined with intense edges), applies some morphological operations and
    • Then the extracted contours are send to the feature extractor where features for each contours are extracted.
    • These features are then classified as license-plate and non-license plate.
    • A high probability indicates a contour to be the license plate.
    • Finally the contoured classified as a license plates (high probability) are stashed in a directory.

    -- Look at the "config-local.conf" to get a understanding of the directory name.