Accurate License Plate Recognition and Super-Resolution Using a Generative Adversarial Networks on Traffic Surveillance Video

Automatic License Plate Recognition (ALPR) is one of the most important methods of intelligent traffic surveillance applications. Some existing ALPR systems are developed for near-frontal plate images in a single lane. However, most surveillance cameras have a challenging environment: small size object, poor resolution and blurred image. We propose a new method that can be applied in the ALPR challenged environments by using super-resolution (SR) module based on Generative Adversarial Networks (GAN). We also used the state-of-the-art and real-time object detection method, You Only Look Once (YOLO), for license plate detection and character recognition. We collected a challenging dataset at low resolution and small object less than 60*60 size and evaluate our approach on it. The achieved mean accuracy of recognition of license plate is above 2% better than other methods in our dataset. Our implementation demonstrate the superiority over the state-of-the-art.