Application of Broad Learning System for Container Number Identification

Abstract

Due to the Information Technologies (ITs) and Computer Technologies (CTs) have been dramatically developed in recent years, harbor cities pay more attentions on implementing the smart ports. In such kind of ports, the containers loading and uploading are almost autonomous, this can sufficiently enhance throughput ability and improve the management efficiency. To address this issue, automatic and correct identification of containers number is the bottleneck, and also the key technology. The container number characters are usually deformed or missing due to influences caused by rain, fog, oil stains, and creases on the surface of the containers, which would influence the recognition accuracy rate. Therefore, this study introduces a novel method named Broad Learning System (BLS) for identification of the container number characters. To compare with other methods, our algorithm presents fast training speed and high testing accuracy, which makes it more suitable for container number identification in practice.