닫기

- 한국법과학회 로그인 화면입니다. -

닫기

논문검색

THE KOREAN SOCIETY OF FORENSIC SCIENCES

HOME 학회지 논문검색
pISSN: 1598-0715

한국법과학회지, Vol.23 no.1 (2022)
pp.16~24

DOI : 10.53051/ksfs.2022.23.1.003

- CNN을 이용한 EES 유효성 검증에 관한 연구 -

백세룡

(가천삼송자동차연구센터 연구원)

김연우

(가천삼송자동차연구센터 연구원)

김천호

(가천삼송자동차연구센터 연구원)

김종혁

(국립과학수사연구원 교통과 감정관)

김종혁

(국립과학수사연구원 교통과 감정관)

There is a function that describing the severity of vehicle’s accident by (changes of speed in the center of gravity of vehicle, km/h), impact point of the collision vehicle, the impacted point, and EES(Energy Equivalent Speed, km/h) using depth of impacted vehicle and the energy loss of the vehicle. CNN(Convolutional Neural Networks) is mainly used in Deep-learning to process image or video data. Convolution is a Neural Network model before preprocessing task. In PC-Crash V 13.0 has a EES-CNN function using EES and CNN. When user inputs image file of damaged vehicle, EES-CNN shows predicted  with a bar graph. To utilize this EES-CNN function in real accident-analysis, it has to be verified or needs improvements. In this study, after collecting images from NHTSA NASS-CDS in frontal collision condition, then verified if predicted data matches to Δυ data from NASS. Furthermore we analyzed property of pictures and presented a method of taking a picture for EES-CNN function to get higher accuracy.

다운로드 (Download) 리스트(List)