In this work, we propose a scheme using deep convolutional neural network (CNN) to detect and recognize vehicle license plates in complex natural scene. In particular, first, we propose to leverage the target detection method which named you only look once (YOLO) based on deep learning to detect the license plates. We optimize the network structure and train a 30-class CNN which can perform real time detection. Next, we combine the advantages of Dense Convolutional Network (DenseNet) and Residual Network (ResNet) and propose a simple, highly efficient network model named RDNet to recognize the license plates. Last, we concatenate two well-trained networks to detect and recognize license plate with high accuracy. The proposed scheme based on deep CNN needs free segmentation and the whole process needs no manual intervention. Extensive experiments verify the effectiveness and robustness of our proposed scheme, and the recognition accuracy achieves 99.34%.
CITATION STYLE
Gao, P., Zeng, Z., & Sun, S. (2019). Segmentation-free vehicle license plate recognition using CNN. In Lecture Notes in Electrical Engineering (Vol. 494, pp. 50–57). Springer Verlag. https://doi.org/10.1007/978-981-13-1733-0_7
Mendeley helps you to discover research relevant for your work.