Regresar

End-to-End License Plate Recognition System for an Efficient Deployment in Surveillance Scenarios

Abstract:

The automatic recognition of license plates has been a widely research area, in special for surveillance scenarios. With the evolution of vigilance cameras and embedded devices, many computer vision models are being used to identify license plates by recognizing its characters. In this work, we present a computer vision model based on convolutional and recurrent neural networks to solve the character recognition task. Artificial data generation is used to balance the dataset, composed by images of alphabetical and numerical characters. A weighted loss function is used in the training stage to enable a robust generalization of the model in the real-world. Our approach reaches an average recognition accuracy of 95% at 35 fps when tested on 60 real surveillance videos.