A road in Vietnam being monitored by ‘AI-based intelligent CCTV technology’ tailored to local traffic conditions for smart city development. Provided by KICT.
A research team from the Korea Institute of Civil Engineering and Building Technology (KICT) has developed an ‘AI-based intelligent CCTV technology’ that reflects Vietnam’s unique road traffic characteristics to aid in the development of the country’s smart cities.
KICT announced on the 11th that the research team, led by Lim Seong-han, Head of the ITS Performance Evaluation Center, has presented a practical solution for localizing smart city technology by developing an AI-based intelligent CCTV system optimized for Vietnam’s road traffic environment.
The Vietnamese government has designated smart cities as a national future growth engine and is pursuing a phased expansion strategy until 2030. Pilot smart city projects are currently underway in major cities such as Hanoi, Da Nang, and Ho Chi Minh City.
In the transportation sector, in particular, the demand for advanced technologies such as AI and big data-based real-time monitoring and incident detection is rapidly increasing. This has brought the need for intelligent CCTV technology, which can detect risk factors like wrong-way driving and traffic accidents in real-time, to the forefront.
The KICT research team developed the technology with an AI deep learning-based algorithm at its core. While large-scale, high-quality training data is essential to enhance detection and judgment accuracy, acquiring such data in Vietnam is challenging, time-consuming, and costly.
To achieve high performance even with the limited local training data, the research team applied the ‘transfer learning’ technique. To supplement the insufficient training video data, they also utilized high-quality data secured in Korea and retrained the algorithm to optimize it for Vietnam’s local traffic conditions.
Based on the algorithm and data, the KICT research team developed a Vietnamese traffic object recognition algorithm, a traffic situation judgment algorithm, and an incident detection algorithm. The developed algorithms were installed in intelligent CCTV prototypes at three locations in a smart city project in Tam Ky City, Quang Nam Province, where field operational tests and performance verifications were conducted.
As Vietnam currently lacks official standards to verify the performance of intelligent CCTV, the performance was reviewed in accordance with the ‘ITS (Intelligent Transport Systems) Performance Evaluation Standards for Automobiles and Road Traffic’ established by Korea’s Ministry of Land, Infrastructure and Transport (MOLIT).
The review demonstrated high performance, with an object recognition accuracy of 99.6% for pedestrians, two-wheelers, and automobiles, and an incident detection accuracy of 97.8%. Through this technological development, KICT has secured a core technology for transportation infrastructure that aligns with Vietnam’s smart city strategy. It is also expected to contribute to strengthening the technological competitiveness of Korean SMEs and laying the foundation for their expansion into overseas markets.
KICT President Park Seon-gyu stated, “This research achievement is a significant contribution to the practical localization of Vietnam’s smart city transportation sector.” He added, “KICT will continue to actively support the overseas expansion of domestic companies and take the lead in global smart city technology.”
The research was developed through the ‘Development and Local Application of Vietnam Smart City-Specific Intelligent CCTV Technology (2023-2024)’ project, a KICT support program for small and medium-sized enterprises funded by the Ministry of Science and ICT. The technology has been transferred to Yungchang.
Copyright ⓒ DongA Science. All rights reserved.
