The Improvement of Character Recognition on ANPR Algorithm using CNN Method with Batch Normalization

Automatic Number Plate Recognition (ANPR) is a highly accurate system that is capable to read vehicle number plates without human intervention which is mainly divided into three steps: plate localization, character segmentation, and character recognition that most affect the accuracy of the ANPR. Convolution Neural Network (CNN) is one of the popular methods which commonly used recently that gives an impressive performance. The method that proposed is to add a layer of batch normalization. To evaluate the proposed method, a dataset consisting of image number plate characters was divided into 36 classes, which represent the letters A–Z and the numbers 0–9. The model was trained using augmented and non-augmented datasets of 1020 for training and 180 for validation. After being trained for 25 epochs, the model was tested using 200 different datasets for image test data. The experimental results show that the proposed method can increase the accuracy from 95% to 98%, which is 3% better.