Intellivix to convert 4,700 highway CCTV cameras to AI surveillance

AI safety company Intellivix will deploy artificial intelligence to detect and respond in real time to hazardous conditions across the country's expressway network.

Intellivix said Thursday it had won a contract for the "public AI closed-circuit television (CCTV) conversion project," supervised by the National Information Society Agency (NIA) with Korea Expressway Corporation as the end-user agency.

The project will convert existing expressway CCTV cameras nationwide into an intelligent AI surveillance system by pairing domestically developed neural processing unit (NPU)-based video analysis servers with visual language model (VLM) servers, without replacing the existing camera hardware.

The project covers a total of 4,700 channels: 3,378 in expressway tunnels nationwide, 422 at sections prone to road icing, 200 at accident-prone stretches, and 700 at heavily congested sections.

The AI safety monitoring system Intellivix will build analyzes CCTV footage in real time, detecting a wide range of road incidents — including traffic accidents, broken-down or stopped vehicles, traffic congestion, wrong-way driving, pedestrians on the road and fallen debris.

Intellivix plans to apply a dual-analysis architecture combining vision AI, which excels at detection, with a VLM capable of understanding footage context — both built on the company's proprietary video analysis core engine, VIX2.0.

In the first stage, vision AI analyzes CCTV footage from expressways nationwide to detect the movement of objects such as vehicles, pedestrians and debris, as well as abnormal situations. In the second stage, the VLM examines footage before and after a detected event, along with the surrounding road environment, to verify whether an actual incident has occurred.

The VLM also automatically generates situation reports for traffic control operators by interpreting the type and location of an incident, the direction of movement of vehicles and pedestrians, lane occupancy and surrounding traffic flow.

According to Intellivix, conventional video analysis systems frequently produced false positives — flagging normal situations as accidents — or missed actual incidents in complex road environments involving vehicle shadows, lighting changes, weather conditions and congested traffic.

Intellivix said the secondary VLM verification step improves the accuracy and reliability of video analysis results, addressing both false positives and missed detections.

All information detected by the AI and verified by the VLM will be managed through Intellivix's next-generation AI safety platform, GenAMS. GenAMS displays the locations of vehicles and pedestrians, incident situation reports, directions of movement and lane occupancy status in real time on a high-definition road map, enabling traffic control operators to grasp the location of an incident and surrounding traffic conditions and take follow-up action.

Intellivix said the project is significant in that it leverages existing CCTV infrastructure while combining domestically made NPU-based video analysis servers with VLM servers. The company expects the project to serve as a case study in AI adoption for public transportation safety — cutting construction costs and time while applying domestic AI semiconductors and AI technology to the sector.

Intellivix added that it has accumulated experience building NPU-based AI video analysis systems, optimizing servers and deploying combined video analysis AI and VLM solutions in the field. The company previously used domestic NPUs to build a wildfire spread prediction system in Hadong-gun and Sancheong-gun, South Gyeongsang Province — a system that uses AI to analyze wildfire surveillance footage along with weather and terrain data to predict the outbreak and spread of wildfires and support disaster response.

Intellivix said it plans to use the project as a springboard to expand its domestically developed NPU-based AI technology — proven in the disaster and safety sector — into transportation, and to advance a nationwide real-time road safety monitoring system.

"This project is significant in that it builds a large-scale AI safety monitoring system using domestic NPUs without replacing existing CCTV infrastructure," Intellivix Chief Executive Choi Eun-su said. "We will contribute to making expressways safer for the public by using AI technology to prevent traffic accidents and secondary collisions."