License Plate Recognition in Wild with Super-Resolution

License plate recognition is an important part of intelligent transport systems. Although there are various license plate recognition algorithms used in commercial activities, most of them have several limitations, requiring high image pixel and low image skew. In this paper, we propose a new license plate recognition system in unconstrained environments, which incorporates plate super-resolution. We use a generative adversarial network for super-resolution(SR) of license plate image to solve the problem of low recognition accuracy due to low-resolution images. Then, the SR image will be input an elaborate designed plate recognition network. We use Chinese City Parking Dataset(CCPD) as the evaluation dataset, on which the experiments show that the proposed method performs better than current license plate recognition methods, especially on low- resolution plate images.