Rail Train Number Recognition Based on Improved VGG-16 Network

As the rail train number recognition system based on deep learning image processing technology was gradually recognized and applied, an improved VGG-16 network was developed for rail train number recognition. The batch standardization (BN) is added to the classical VGG-16 network, and a rail train number character recognition method is designed. Combining with the train number images taken in the Guangzhou Metro, Nanjing Metro, and the laboratory, the train number recognition algorithm is trained and tested. The experimental results show that the comprehensive accuracy rate of the developed method for rail train number recognition reaches 99.54%, which meets the requirements of on-site use.