AI-Integrated Customs Video Management Solution

Overview

The explosive growth of overseas direct purchases has led to a surge in X-ray image data at customs. This has raised concerns regarding decreased accuracy in customs clearance due to the accumulated fatigue of human inspectors using traditional visual methods. Furthermore, missed images have caused persistent data inconsistency issues. To ensure the reliability and speed of the clearance process, the Korea Customs Service has been supporting an AI specialist consortium since 2023 to develop and demonstrate AI solutions for the automatic separation, tracking, and searching of X-ray images.

Project description

The AI-integrated customs image management solution was developed based on large-scale X-ray imagery and cargo declaration data held by the Korea Customs Service. Over 11 million X-ray data points were processed and utilized to develop and validate AI solutions focused on three core functions: image separation/storage, tracking/merging, and DB search.

This AI solution encompasses a variety of sophisticated features. The Image Separation & Storage Service automatically disentangles overlapping items in X-ray views, matching them with declaration info for visual clarity. The Image Tracking & Merging Service integrates these separated images with AI diagnostic results into a single, real-time interface. Additionally, the Image DB Search Service enables rapid identification of similar cargo from tens of millions of archived records. Finally, the Alert Service utilizes AI trained on hazardous materials to detect similar shapes in real-time, immediately warning inspectors of potential threats.

Key technological features

• Image Separation & Storage Service: AI automatically separates overlapping items within X-ray images and visually organizes them by matching each item with its declared cargo information
• Image Tracking & Merging Service: Provides a real-time integrated view on a single screen, combining separated image data with AI analysis results
• Image DB Search Service: High-speed AI-driven search to identify similar cargo among tens of millions of X-ray images stored in the existing database
• Real-time Alert Service: AI trained on hazardous item imagery identifies similar shapes in real-time and issues warnings to users upon detection

Expected outcomes

The AI-based customs image management solution has undergone continuous refinement through field demonstrations at major customs sites, including the Incheon International Airport Customs Express Cargo Logistics Center and Pyeongtaek Customs. As a result, customs inspection accuracy has improved to over 95% compared to manual visual inspections, with image similarity searches achieving an average accuracy of 96.7%.

Notably, the time required for cargo matching and image searching has been reduced from several minutes to under 1 and 2 seconds, respectively. This has led to a 26-fold increase in customs inspection speed (from 5 cases/min to 130 cases/min). This performance boost is expected to lower the fatigue of customs personnel, significantly reduce human error, and facilitate safer and faster clearance. Furthermore, by proactively blocking the entry of hazardous items such as narcotics, the system is anticipated to secure both national security and public safety simultaneously.

Launch year: 2023

Attribution

Case study courtesy of the National Information Society Agency, Republic of Korea, Apolitical's Content Partner. Sourced from Use Cases of Public AI Service, Vol.2, https://eng.nia.or.kr/site/nia_eng/ex/bbs/View.do?cbIdx=31975&bcIdx=29182&parentSeq=29226