Automatic real-time identification of international mail: system utilizing AI for X-ray image recognition and processing

In order to improve the sophistication and efficiency of customs inspections, NEC received an order from Japan Customs for an X-ray image recognition and processing system that automatically identifies international mail subject to inspection in real time using artificial intelligence (AI). The plan is to implement the system from April this year at the Tokyo overseas mail sub branch handling customs procedures for international mail.

Cross-border e-commerce (EC) cargo and personal imports that accompany EC have been increasing with the spread of the Internet, resulting in a global increase in the movement of goods. Under such circumstances, prompt customs clearance procedures for international mail are required, and strict border control has also become increasingly important. Until now, customs staff have manually inspected international mail using X-ray inspection equipment to verify that the packages do not contain prohibited goods or regulated imports--for example, by checking the images themselves. However, since the volume of international mail and, as a result, the number of inspection operations are expected to rise in the future, there is a need for a new sophisticated and efficient system to handle this.

The system that NEC received an order for this time has a high-precision AI learning model that it has jointly developed with the Japanese Ministry of Finance using RAPID machine learning--it is part of NEC's cutting-edge AI technology lineup NEC the WISE. When international mail goes through the X-ray inspection device, AI analyzes and identifies the X-ray images in real time and then conducts primary screening of the item(s) in question. This allows for more sophisticated and efficient inspections while reducing the workload of the customs staff. Furthermore, it is equipped with a routine accuracy monitoring function to maintain AI accuracy, thus making it possible to maintain and improve the learning model so that AI accuracy does not decline even if mail contents change. This system will be implemented as part of measures set forth by the Customs Bureau of the Ministry of Finance in June 2020 in the "SMART Customs Initiative 2020", a medium-to long-term vision of customs administration.