Vehicle License Plate Recognition System Based on Deep Learning Deployed to PYNQ

In this paper, we propose a Vehicle License Plate Recognition System (VLPRS) based on deep learning deployed to Xilinx PYNQ board, which includes sequentially license plate detection and license plate recognition. We compare the efficiency and precision of YOLOv2 and YOLOv3 in detecting license plates from original images. The recognition part applies Convolutional Neural Network (CNN). We implement license plate recognition on PYNQ. Considering of the hardware limitation, the parameters we use in the license plate recognition are as few as possible. Conducted on the images captured by cameras in wild environment, the experiment is convincing and the results demonstrate that VLPRS performs well. The detection part achieves the precision of 99.35% with the speed of 12. 19ms per image while the per-character accuracy of recognition part is above 97.89% with few parameters. The recognition on PYNQ achieves the same accuracy.