LaonPeople Selected for KRW 2.4 Billion Government Project to Develop “Coastal Safety AI Response Solution”

VLM-Based AI Video Analysis and Automated Response Guidelines to Strengthen Coastal Safety

LaonPeople has been selected as the lead organization for a KRW 2.4 billion government-funded project to develop an AI-powered coastal risk-response solution.

AI technology company LaonPeople (CEO Seok-Joong Lee) announced on the 9th that it has been chosen as the primary contractor for an AI development and verification program supported by the National IT Industry Promotion Agency (NIPA). The company will work in consortium with Mobilint, an AI semiconductor design fabless firm, to execute the project over the next two years.

The initiative aims to prevent accidents along Korea’s coastline and ensure rapid rescue and response by developing an advanced VLM (Visual Language Model)-based coastal video-analysis AI solution. In collaboration with the Korea Coast Guard and other related agencies, the project will establish an intelligent monitoring system capable of detecting coastal incidents in real time and supporting immediate response actions. By combining VLM video-analysis technology with RAG (Retrieval-Augmented Generation)-enhanced LLM (Large Language Model) automated response-guideline capabilities, the system will be designed to identify risks and initiate response measures within 30 seconds of an incident.

The project, scheduled to run through 2026, will first focus on training object-detection and semantic-segmentation models using a wide range of datasets, including Coast Guard video archives, new CCTV footage, and open data sources. The team will also implement a VLM-based situation-description model. In designated pilot areas, the system will be optimized for incident detection and automated reporting using fixed CCTV data, followed by verification through a KOLAS-accredited testing agency.

During the second year, the project will enhance performance in low-visibility environments, improve VLM accuracy through DPO (Direct Preference Optimization) training, and build a fully automated RAG-LLM response-guideline system. The team will also design an active surveillance framework that integrates mobile imaging equipment such as drones—addressing the limitations of relying solely on fixed CCTV cameras.

A LaonPeople spokesperson stated, “This project will help eliminate blind spots in coastal surveillance and secure the golden time for rescue operations, significantly reducing the risk of casualties.” The spokesperson added, “By advancing and integrating these technologies, we plan to expand the system’s application to various disaster-response environments, including remote islands, underground tunnel flooding, and wildfires.”