CHOSUN

주야간 사고예방을 위한 차량용 블랙박스 시스템에 관한 연구

Metadata Downloads
Author(s)
김강효
Issued Date
2016
Keyword
블랙박스 주야간 사고예방
Abstract
A car black box is an image recorder used to identify the cause of a car accident by recording information, for example, images, time and a volume of impact at the time of the accident. In most car accidents, it is hard to get enough evidences at the site of an accident to result in difficulty in accurate analysis of the accident. Although statement by witnesses is very important when analyzing a car accident, it is hard to find witnesses in most cases. Because a car black box device saves information about situations that occur while a car in question with the on-board black box travels to allow analysis of car accidents and responsibility thereof, using car black boxes is increasingly common now.
Recently, an intelligent black box is studied, which implements self-decision and copes with dangerous situations in real time as people are more interested in technology and safety. Exemplary intelligent black box technologies being studied at present include accident prevent systems for protecting drivers from accidents that can occur while travelling, for example, the lane departure sensing system and the forward collision warning system.
However, the aforementioned systems being studied cannot prevent incidents in advance, for example, theft, stealing or run after causing physical damages by malicious human behaviors when a car in question stops or is parked. In addition, it is essential that a user must examine all saved images when an accident occurs. Another inconvenience is that it is hard to get sharp images captured at night of low illumination.
Therefore, this thesis aims to suggest a system for correcting disadvantages of conventional black box systems and preventing incidents by malicious human behaviors when a car in question stops or is parked at night or during the day. The suggested car black box system uses the depth camera of a Kinect sensor for predicting dangers to detect persons and measure distances, and sets up three-step areas of interest to examine the level of person's approach. In addition, it uses the CIE L*a*b* color space to examine whether it is the daytime or nighttime. After the process, it generalizes the result of detected persons in the areas of interest and the daytime or nighttime to prevent incidents while the car in question stops or is parked to create an accident prevention event compatible with a daytime situation and a nighttime situation.
The result of an experiment reveals that the suggested car black box system normally provides signals of detected persons and warnings when they approach the car in question in outdoor parking lots during the day, indoor parking lots during the day, outdoor parking lots in the early evening, indoor parking lots in the early evening, and outdoor parking lots at night. Therefore, it is proved that the suggested system can prevent incidents by malicious human behaviors while a car in question stops or is parked at any time at night or during the day.
Alternative Title
A Study on the Vehicle Black Box System for the Day and Night Accident Prevention
Alternative Author(s)
Kim, KangHyo
Affiliation
조선대학교 산업기술융합대학원
Department
산업기술융합대학원 소프트웨어융합공학과
Advisor
반성범
Awarded Date
2016. 2
Table Of Contents
제1장 서 론 1
제1절 연구 배경 1
제2절 연구 목적 및 방법 4
제2장 지능형 블랙박스 시스템에 대한 관련 연구 6
제1절 사람 검출 6
제2절 조도 판단 12
제3장 주야간 사고예방을 위한 블랙박스 15
제1절 키넥트를 이용한 사람 검출 18
1. 키넥트 센서의 특징 18
2. 키넥트 SDK 19
3. 사람 검출 및 거리 측정 20
제2절 CIE L*a*b* 색상 정보을 이용한 주야간 판단 22
1. CIE L*a*b* 색상 공간 22
2. 주야간 판단 23
제3절 거리에 따른 관심 영역의 설정 24
제4절 사고예방 이벤트 25
제4장 실험 및 결과 28
제1절 실험 환경 28
제2절 실험 및 결과 28
제3절 제안하는 시스템의 성능 분석 및 비교 35
1. 사람 검출률 35
2. 주야간 판단률 40
3. 기존의 블랙박스 시스템과의 성능 비교 42
제5장 결 론 45
참고문헌 47
Degree
Master
Publisher
조선대학교 산업대학원
Citation
김강효. (2016). 주야간 사고예방을 위한 차량용 블랙박스 시스템에 관한 연구
Type
Dissertation
URI
https://oak.chosun.ac.kr/handle/2020.oak/1946
http://chosun.dcollection.net/common/orgView/200000265481
Appears in Collections:
Engineering > Theses(Master)(산업기술창업대학원)
Authorize & License
  • AuthorizeOpen
Files in This Item:

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.