ETRI-Knowledge Sharing Plaform

ENGLISH

성과물

논문 검색
구분 SCI
연도 ~ 키워드

상세정보

학술대회 Fast and energy-efficient object detection and tracking for de-identification in video
Cited 0 time in scopus Download 2 time Share share facebook twitter linkedin kakaostory
저자
이준혁, 이승재
발행일
202210
출처
International Conference on Consumer Electronics (ICCE) 2022 : Asia, pp.404-405
DOI
https://dx.doi.org/10.1109/ICCE-Asia57006.2022.9954651
협약과제
22IH2600, AI 기반 전통 건축 손도면 CAD 도면화 기술 개발, 이승재
초록
In modern life, camera modules are available everywhere, including smartphones and home appliances so efficient de-identification methods are continuously required to protect personal information. In this paper, we propose object detection and tracking method that operates within the decoding process in video for fast de-identification. We reduce computational complexity by partially detecting personal information such as faces and vehicle license plates in GoPs and tracking locations using object displacement features. The proposed method confirms the reduction in computation and power consumption through experiments.
KSP 제안 키워드
Computational complexity, De-Identification, Identification method, Object Detection and Tracking, Personal information, Power Consumption, Tracking method, energy-efficient, home appliances, license plate, vehicle license