Turkish Vehicle License Plate Recognition Using Deep Learning

Abstract

License plate recognition systems are commonly used in various real-life applications. Traditionally, several image processing techniques have been used in these systems. Although these solutions provide adequate results in general, in some cases incorrect results are obtained. In this study we use Deep learning to provide a highly accurate license plate recognition system. We also employ image processing steps in the process to increase the accuracy of the results from convolutional neural. The system uses TensorFlow framework in the background and developed using Keras deep learning library. We created a new Turkish vehicle license plate dataset for this study using real world vehicle images obtained from Firat University security cameras. The results show that the system performs well and the trained neural network can highly accurately recognizes the license plates.