인문학
사회과학
자연과학
공학
의약학
농수해양학
예술체육학
복합학
지원사업
학술연구/단체지원/교육 등 연구자 활동을 지속하도록 DBpia가 지원하고 있어요.
커뮤니티
연구자들이 자신의 연구와 전문성을 널리 알리고, 새로운 협력의 기회를 만들 수 있는 네트워킹 공간이에요.
논문 기본 정보
- 자료유형
- 학술저널
- 저자정보
- 발행연도
- 2026.6
- 수록면
- 791 - 796 (6page)
- DOI
- 10.5302/J.ICROS.2026.26.0060
이용수
초록· 키워드
The difficulty of acquiring repeatable datasets in disaster environments, such as smoke and darkness, hinders the systematic evaluation of perception systems for high-assurance unmanned mobility. This study proposes Unity’s high-definition-render-pipeline-based simulation framework that parameterizes disaster factors—including fire location, smoke level, and heat sources—to generate repeatable scenarios and synchronized multimodal observations. The simulator integrates volumetric smoke for visibility degradation, a lightweight thermal propagation model, and a configurable thermal camera model. Using the You Only Look Once, version 8 model, we evaluated person detection performance on simulated red–green–blue (RGB) and thermal images across varying smoke levels. To examine whether the simulation-observed modality trend persists under severe visible-light degradation conditions, real-world experiments were conducted under normal and blackout conditions using RGB (RealSense D455) and thermal (Optris Xi 400) cameras. The results consistently showed that RGB-based detection degraded abruptly in dense smoke or darkness conditions, whereas thermal-based detection remained comparatively stable. These findings suggest that the proposed simulator can serve as a controlled tool for analyzing modality-specific detection robustness under visibility-degraded conditions.
상세정보 수정요청해당 페이지 내 제목·저자·목차·페이지정보가 잘못된 경우 알려주세요!
목차
- Abstract
- I. 서론
- II. 제안 방법
- III. 실험 설정 및 평가 지표
- IV. 실험 결과
- V. 결론
- REFERENCES