한국전자통신연구원(ETRI)이 자체 개발한 시각 인공지능 기술인 ‘딥뷰(DeepView)’기반 CCTV를 활용한 현수막 탐지 기술개발 실증에 경남 사천시와 경북 경산시에 우선 적용한다.
.JPG)
▲ETRI researchers are demonstrating deep view technology.
Public CCTV utilization for citizen safety and local government budget reduction
A domestic research team announced that they will fully utilize visual intelligence technology to detect illegal banners and analyze their contents using public CCTV.
The Electronics and Telecommunications Research Institute (ETRI) announced on the 5th that it will first apply its self-developed visual artificial intelligence technology, ‘DeepView’, to Sacheon-si, Gyeongsangnam-do and Gyeongsan-si, Gyeongsangbuk-do for demonstration of banner detection technology using CCTV.
The research team decided to fully cooperate with the Korea Local Finance Corporation, Sacheon City, and Gyeongsan City under the Ministry of the Interior and Safety, which oversees outdoor advertising projects for local governments across the country, to conduct a demonstration project to detect illegal banners.
ETRI focused on developing technology that can operate in real environments through close collaboration with the Korea Outdoor Advertising Center within the Korea Local Finance Corporation on functions, requirements, and data necessary for detecting illegal banners.
Illegal banners installed indiscriminately in school zones, intersections, and four-way intersections have not only damaged the city's appearance, but have also threatened the safety of citizens and frequently caused inconvenience and complaints.
On the other hand, there was a problem that immediate response was difficult due to lack of local government manpower and budget, and that the focus was on post-response.
In addition, the national budget support for outdoor advertisement removal, compensation system, maintenance services, etc., and the budget required by local governments are increasing every year.
In particular, the current political partyIn the case of water curtains, there is no reporting obligation or location restriction, so they are installed in droves at intersections, etc., causing great inconvenience to citizens.
According to data from the Ministry of Environment, the weight of political party banners reported to have been removed by local governments across the country in the first quarter of this year exceeded 1,300 tons, and most of them were illegal banners that had been hung for more than 15 days.
In addition, in the case of political party banners, the posting date is too short compared to the posted text, making it difficult to confirm the posting deadline, so it is very important to determine the time of installation.
Therefore, in order to determine whether or not it is illegal, it is necessary to check the installation time based on the installation location area and character recognition results along with the act of installing the banner.
To this end, the research team developed a method to detect the banner installation area using deep learning-based artificial intelligence technology, as well as recognize the installation time and content of the banner through character detection that is robust to changes in letter shape and character recognition using a large-scale visual language model.
In addition, the technology is expected to significantly improve the work efficiency of local governments by detecting illegal banners in the city center and computerizing the content while linking with local government CCTV video control systems.
In particular, it is expected to greatly contribute to preventing the installation of illegal banners in advance, preventing citizen safety accidents, and reducing local government budgets while reducing budget and administrative power.
The research team plans to first apply this technology to Sacheon City and Gyeongsan City to verify the technology, and then work with the Korea Local Finance Corporation to support the expansion and dissemination of related technology to local governments across the country.
'DeepView', a visual intelligence technology developed by ETRI, is a core source technology for visual intelligence that deeply analyzes large-scale images and videos and applies them to various environments.
Applied to solve various social problems, from preventing safety accidents in the city center to cracking down on illegal dumping of wasteThe results have been achieved.
This has greatly contributed to enhancing the competitiveness of related industries.
The ETRI Visual Intelligence Lab research team has been recognized for their excellent research results, including being selected as ETRI's representative achievement in 2020 and as one of the 100 best national research and development achievements in 2019, 2021, and 2023.
Our research team is currently carrying out the '(Deep View) High-Performance Visual Discovery Platform Development Project for Real-Time Large-Scale Image Data Understanding and Prediction' as part of the Innovation Growth Engine Project (Artificial Intelligence) supported by the Ministry of Science and ICT and the Information and Communications Technology Promotion Center (IITP).