A Method of Bank Card Number Recognition and End-to-End Integration Based on Deep Learning

Accurate identification of bank card number has a wide range of application scenarios in banking business. Because of the complexity of data samples, the lack of data sets and the confidentiality of information, the identification of bank card numbers is still a difficult task. In this paper, we present a bank card Network (BCNet) to identify bank card numbers and integrate an end-to-end system which can be used conveniently in banking business. It includes two core architectures: text detection and recognition. In order to adapt to various bank card number recognition scenarios, we optimize the data enhancement and proofread the recognition results. Experiments show that the recognition accuracy of our model reaches 99%, and the system takes 58ms to run. The system shows excellent performance in both accuracy and running time.