Abstract
There are many types of vehicle license plates in China, including new energy license plates, large truck license plates, government vehicle license plates, and military license plates. The existing commercial license plate recognition system only targets common license plates and does not completely cover the full range of license plates. Therefore, this paper proposes an SSD-based end-to-end license plate recognition system (LPR-SSD). The LPRSSD network architecture consists of upper and lower classification networks: the upper layer network is used for vehicle license detection and classification, and the lower layer network is used for license plate character detection and classification. In order to enhance the generalization performance of the LPR-SSD network, in addition to the real license plate image captured by the camera, this paper synthesizes 50K simulated license plates for each type of license plate according to the legal document [1]. Experiments show that LPR-SSD achieved a faster convergence speed during training. After the test set verification, the accuracy of license plate location detection and classification reaches 98.3%, and the character recognition accuracy rate reaches 99.1%.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have