http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
드론 활용 교량 안전점검을 위한 딥러닝 균열 분석에 관한 연구
최대영(Daeyoung Choi),백승현(Seung Hyun Paik),김영규(Young-Kyu Kim),정상우(SangWoo Jung),김대년(Dae-Nyeon Kim) 한국정보기술학회 2021 한국정보기술학회논문지 Vol.19 No.12
Visual-safety inspection and investigation of existing bridges have problems with low objectivity and reliability because of the inspectors subjectivity. As the visual inspection technology for deteriorated infrastructure becomes important, research of inspection technology using drones has been proposed. In this study, We propose an alternative method to detect and analyze crack using drones. The drone is equipped with a special camera and gymbal in the forward and upward directions. For crack analysis, the study was conducted in two layers: extraction and analysis. Crack is segmented by learned U-NET model in real time and the Intersection-over-Union (IoU) value was 0.9125 (91%). After, through post-processing image processing of 8 steps, the crack is isolated, clustered, analyzed and visualized.
Diamond ArUco 마커를 이용한 드론 착륙 지점 3차원 좌표 추정 방법 연구
최대영(Daeyoung Choi),백승현(Seung Hyun Paik),김영규(Young-Kyu Kim),유종호(Jong-Ho Yoo),정상우(SangWoo Jung),김태환(Tae-Hwan Kim),김대년(Dae-Nyeon Kim) 한국정보기술학회 2022 한국정보기술학회논문지 Vol.20 No.12
As interest in autonomous landing of drones and unmanned systems is rising, research for accurate landing of drones remains a challenge. In general, there is a landing method through GNSS, but the position error of the landing site is not detailed, so landing methods using vision through a camera and additional sensors are suggested. In the previous study of our research team, CNN-based station recognition research was performed, but the recognition error was limited due to external environment(wind pressure, vortex, etc.) during the landing mission. was presented. The existing identifier is a single identifier, and only one fiducial marker recognized by the drone was relied on. Diamond ArUco, the marker presented in this study, has four identifiers in the form of a cross, so a method to minimize the error of 3D coordinate estimation is presented in this paper. evaluation was discussed.
회전익 드론 접촉식 충전스테이션 도킹을 위한 딥러닝 기반 자동착륙시스템 구현
김영규(Young-Kyu Kim),최대영(Daeyoung Choi),백승현(Seung Hyun Paik),정상우(SangWoo Jung),김대년(Dae-Nyeon Kim) 한국정보기술학회 2020 한국정보기술학회논문지 Vol.18 No.10
Rotary wing drones (RWDs), which are a kind of unmanned aerial vehicles, have been researched and developed to employ various fields, such as logistics transport, observation, surveillance, and measurement, because drones are able to overcome most restrictions resulted from constructions, roads, and geographical features of ground. However, RWDs have a short flight time caused battery issues, and the flight time becomes a major setback for those work. To relieve battery issues of RWDs, various charging stations and its management for RWDs have been proposed. Therefore, this paper proposes and implements a precise automatic landing system in order to accurately dock on a charging port in contact type charging stations for accomplishing missions of RWDs. The implemented automatic landing system uses FC-HarDNet based object detection algorithm and pixel-based distance computation for real-time recognition and tracking of the drone charging stations, and helps exact auto-landing for charging battery of drones. In experimental results, the landing distance error of proposed automatic landing system decreased by around 54.59% as compared with previous researches.