메뉴 건너뛰기
소속 기관 / 학교 인증
인증하면 논문, 학술자료 등을  무료로 열람할 수 있어요.
한국대학교, 누리자동차, 시립도서관 등 나의 기관을 확인해보세요
(국내 대학 90% 이상 구독 중)
고객센터 ENG
주제분류

논문 기본 정보

저자정보
(고려대학교) (고려대학교) (고려대학교) (고려대학교) (인포벨리코리아)
저널정보
한국멀티미디어학회 멀티미디어학회논문지 멀티미디어학회논문지 제26권 제11호
오류 신고하기

피인용 0

검색

    초록·키워드

    In the field of agriculture, autonomous tractors have played a significant role in crop management and the implementation of efficient automation systems. Achieving full autonomy in these tractors poses a crucial challenge, notably in accurately detecting person in safety-critical situations. Detecting fallen person is vital for preventing potentially fatal accidents. However, a scarcity of data related to fallen person has raised concerns about overfitting. To address this issue, this paper introduces a new data augmentation method for identifying fallen person using a Automated RoI(Region of Interest) Copy-Paste method, based on Intersection over Union(IoU). The Automated RoI Copy-Paste method uses IoU to automatically segment and set the RoI areas. Subsequently, it composites the segmented masked images within those areas. Furthermore, it utilizes YOLOv8x-seg and Grounded-SAM models to generate masks for fallen person and vehicle bonnets. When comparing the experimental results using the object detection models YOLOv7-tiny and YOLOv8n on the NREC dataset, an accuracy improvement of 1.3% and 10.0% was observed compared to the conventional segmentation Copy-Paste method.

    최근 본 자료 전체보기