Incheon Airport Validates Passenger Flow Management with Superb AI’s Next-Generation Video Analytics Enhanced by NVIDIA AI

Key Highlights

1. Real-time NVIDIA AI Blueprint for video search and summarization (VSS) powered by agentic AI —Implementing digital twin for airport safety
2. AI and digital twin technology leveraged to process vast amounts of video data across the airport—demonstrating the potential to enhance passenger experience through efficient operations and real-time analysis.
3. NVIDIA AI Blueprint for VSS enables efficient management of large-scale video data—transforming airport safety operations with integrated real-time and historical video search

Superb AI, a leader in end-to-end MLOps for intelligent video analytics, has successfully completed a proof-of-concept (PoC) with Incheon International Airport using its next-generation solution leveraging visual AI agents. This PoC on passenger flow management combines Incheon Airport’s logistics expertise and integrated data design capabilities with Superb AI’s vision language model (VLM) technology and the NVIDIA Jetson platform, NVIDIA TensorRT libraries, and agentic capabilities accelerated by the NVIDIA AI Blueprint for video search and summarization (VSS). The Superb AI platform demonstrated its ability to detect and analyze a wide range of scenarios in real time across the airport’s vast infrastructure, proving its potential to significantly improve operational efficiency and enhance the passenger experience by shortening wait times.

Challenge: The Need for Transformational Solutions

South Korea’s Incheon International Airport operates in a complex environment where multiple entities manage different parts of the facility, making it difficult to maintain a unified and comprehensive operational system. 2024 brought a series of critical challenges for Incheon Airport: the limitations of existing IoT sensors and software, rising passenger complaints due to longer wait times at departure areas, and mounting infrastructure and staffing constraints driven by terminal expansion. To address these multifaceted issues, Incheon airport designated this as the first year of ‘Digital Airport Innovation,’ and urgently required an innovative, AI-driven solution. To explore the potential of technology-driven innovation in airport operations, Incheon Airport selected Superb AI’s Agentic AI solution.

Solution: Breaking Through Legacy Barriers with a Fusion of Next-Generation Technologies

Superb AI introduced a breakthrough solution that redefined what’s possible in video monitoring and analytics. The tested platform was built with Agentic AI, a fusion of Superb AI’s Video Analytics solution and NVIDIA AI Blueprint for VSS.

Thousands of cameras across the airport capture video in real time, generating hundreds of hours of footage each day. Manually reviewing vast volumes of video is inefficient and risks missing critical moments. Superb AI’s intelligent video analytics solution helped solve this by automating the observation process and using AI to identify and classify objects across camera feeds in real time. To make this data actionable and better support safety and security teams in making rapid decisions, Superb AI’s solution leverages VLM and multimodal AI to create a visual AI agent capable of intelligent video summarization. Based on Incheon Airport’s extensive video data, the platform identifies key frames, removes redundant or irrelevant segments, and generates a context-aware summary that preserves the flow of events—enabling rapid situational understanding for security personnel. These AI-generated summaries can be searched using natural language, allowing security operators to instantly identify and respond to incidents.

Natural language search enables operators to retrieve the required information through intuitive queries. Prompts such as “Show congestion at Terminal 1 departures” immediately return precise, relevant footage. The adoption of digital twin technology also enables the visualization of real-time, airport-wide monitoring in a virtual environment while supporting predictive analytics. The system further improves video analysis accuracy by combining advanced technologies such as RGB cameras, object permanence, movement analysis, and zero-shot object recognition.

Incheon International Airport, amid a complex operational environment, faced challenges such as the limitations of existing IoT sensors and software, increasing passenger wait times, facility expansion, and staffing shortages. To address these issues, the airport adopted the innovative AI-based solution Agentic AI (VA-VSS), combining its operational expertise with the technical collaboration between Superb AI and NVIDIA—and successfully validated tangible results.

RESULTS

• Delivering Ultra-Low-Latency Real-Time Video Analysis - By harnessing the NVIDIA Jetson platform’s powerful GPU acceleration and TensorRT’s optimized inference capabilities at the edge, Superb AI ingests large-scale video data with ultra-low latency. This has significantly improved the speed and accuracy of movement analysis, abnormal behavior detection, and emergency response.

• Validating Digital Twin Implementation Across a Large Airport with Multi-Camera Video Streams - Video streams collected from numerous cameras were analyzed simultaneously and visualized on the airport’s map, enabling an accurate replica of physical spaces in a digital environment—enhancing operational efficiency and strengthening predictive capabilities.

• Detecting Anomalies and Risks Using VLMs - By leveraging Vision Language Models (VLMs), the system extracted key scenes from massive video datasets and accurately identified abnormal behavior and risk scenarios, supporting rapid incident response.

• Validating the Potential of Movement Analysis using the Re-ID Technology - Re-identification (Re-ID) technology enabled precise tracking and analysis of movements and travel times. This not only improved congestion management and passenger flow efficiency, but also demonstrated the potential for quick response to someone missing, potentially saving someone who is delayed due to a medical incident or above average security wait times.

Benefit: Creating New Value Through Collaborative Synergy

The integration of the NVIDIA AI platform and Superb AI’s software, in collaboration with Incheon Airport’s logistics and data infrastructure expertise, has produced far more than a mere technical upgrade—it has validated a transformative solution that redefines airport operations. Superb AI’s Agentic AI system continuously learns and evolves to meet the specific conditions and needs of Incheon Airport, driving ongoing operational excellence.

Accuracy Validation Results

• Maintained a mean absolute error (MAE) of just 2–3 people even under high congestion, with further improvement potential through optimized camera placement and algorithm fine-tuning
• Re-ID technology achieved the MAE of approximately 2 minutes when calculating travel times, offering valuable insights for improving and optimizing departure area operations

The system has significantly improved congestion analysis across key airport zones such as check-in counters and departure areas, while dramatically increasing the accuracy of passenger wait time predictions. Additionally, the implementation of the Digital Twin technology enabled high-quality simulation of the airport’s physical environment in a virtual space, allowing pre-verification of operational scenarios.

Most notably, the combined technological capabilities of Superb AI and NVIDIA have helped overcome two major limitations of traditional video analytics systems:

1. Customized adaptability: Unlike conventional static AI models, Superb AI’s Agentic AI continuously learns and evolves based on the unique operational context of Incheon Airport
2. Scalable video data analysis: By integrating VLMs, LLMs, NVIDIA NIM, and vector-based Retrieval-Augmented Generation (RAG), the system can efficiently analyze vast volumes of video data through natural language queries

Data-Driven Innovation Opens a New Chapter in Airport Services

The agentic AI system, developed by Incheon Airport and Superb AI with support from NVIDIA, has opened up new possibilities for revolutionizing operational efficiency and enhancing the passenger experience at Incheon International Airport. In addition to improving passenger satisfaction through real-time wait time visibility, the system enables significant cost savings and maximizes operational efficiency through optimal workforce allocation.

Hyun Kim, CEO of Superb AI, stated, “Our AI-powered video analytics technology deployed at Incheon Airport demonstrates high accuracy even under heavy congestion. It will provide critical data to optimize the airport’s logistics and operational processes. In particular, the integration of NVIDIA’s technology with our agentic AI has validated the system’s ability to recognize not only abnormal behavior, but also activities and logistics anomalies—laying the technological foundation for a safer, smarter airport.”

Incheon International Airport Corporation added, “Through this technical validation, Incheon International Airport was able to confirm the potential of advanced AI solutions, such as Agentic AI with Superb AI and NVIDIA, to overcome operational challenges and deliver real-time video insights for enhanced passenger safety and efficiency. Moving forward, we will actively adopt a wide range of innovative technologies, like Vision Language Models and Re-ID, to set new standards in airport security, congestion management, and rapid incident response, ensuring a world-class travel experience for all.”

With the implementation of the system, Incheon Airport has further strengthened its global status as a leading smart airport. In the long term, the airport aims to offer even more diverse and advanced services—such as congestion prediction, abnormal behavior detection, and personalized service delivery.