메뉴 건너뛰기
.. 내서재 .. 알림
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)
로그인 회원가입 고객센터 ENG
주제분류

추천
검색
질문

논문 기본 정보

자료유형
학술대회자료
저자정보
김지송 (한양대학교) 성민재 (한양대학교) 최준원 (한양대학교)
저널정보
한국자동차공학회 한국자동차공학회 춘계학술대회 2021 한국자동차공학회 춘계학술대회
발행연도
2021.6
수록면
951 - 956 (6page)

이용수

표지
📌
연구주제
📖
연구배경
🔬
연구방법
🏆
연구결과
AI에게 요청하기
추천
검색
질문

초록· 키워드

오류제보하기
In this paper, we propose a Deep learning based 2D object detection network and pedestrian-specification network using port CCTV image data. CCTV is installed in a high place and can observe a wide area, so there is almost no occlusion problem and has overall information about the surrounding environment. This point can be applied to security systems for public facilities such as ports and airports, and useful information that cannot be obtained from sensors attached to the vehicle in current autonomous vehicles can be provided to vehicles and pedestrians. In this paper, we developed a model for detecting multiple objects and specifying pedestrians by using port CCTV image data. The proposed method is 1) detecting 2D multi-objects through a GFL (Generalize Focal Loss) network that uses a port CCTV image as an input. The detection performance was improved and generalization error was reduced by applying a data augmentation to the input image that trains the multi-object detection network. In addition, when the detection result is a human object, the detection information is transmitted to a subsequent pose estimation network. The next proposed technique is 2) pedestrian-specification. Key points of a person are extracted through a pose estimation network using human detection information as input. Using AlphaPose as a pose estimation network, extract the overall area of the upper body of the person and then perform a pedestrian specification that obtains the color of the image as an RGB value. The color extraction accuracy is improved with an algorithm that removes a part of a person’s skin color that may appear on the upper arm or the like from the upper region. Finally the average value is calculated for the rest of the image and the color of the image is obtained as an RGB value. To verify the performance of the technique proposed in this paper, we experimented with the port CCTV dataset and the COCO dataset that widely used for 2D object detection.

목차

Abstract
1. 서론
2. 관련 연구
3. 물체 검출 및 보행자 특정화 기법
4. 실험
5. 결론
References

참고문헌 (0)

참고문헌 신청

함께 읽어보면 좋을 논문

논문 유사도에 따라 DBpia 가 추천하는 논문입니다. 함께 보면 좋을 연관 논문을 확인해보세요!

이 논문의 저자 정보

이 논문과 함께 이용한 논문

최근 본 자료

전체보기

댓글(0)

0