Publications

Detailed Information

Developing a Multi-Zone People Counting Methodology using Surveillance Cameras for Search and Rescue Efforts during Building Disasters : 건물재난 탐색구조를 위한 다중 CCTV 기반 인원 계수 방법론

DC Field Value Language
dc.contributor.advisor지석호-
dc.contributor.author박찬혁-
dc.date.accessioned2021-11-30T01:53:06Z-
dc.date.available2021-11-30T01:53:06Z-
dc.date.issued2021-02-
dc.identifier.other000000163809-
dc.identifier.urihttps://hdl.handle.net/10371/175077-
dc.identifier.urihttps://dcollection.snu.ac.kr/common/orgView/000000163809ko_KR
dc.description학위논문 (석사) -- 서울대학교 대학원 : 공과대학 건설환경공학부, 2021. 2. 지석호.-
dc.description.abstractBuilding disasters have been consistently occurring in the past decade. As search and rescue efforts are conducted on the basis of limited information, it is important to count people before the building disaster occurs for quicker search and rescue efforts.
The need for acquiring occupancy information beforehand, has been issued and researched under the field of people counting. However, the scope of previous researches were limited to single or few regions at most due to challenges of installing sensors throughout the building and concerns for replicability. To overcome this deficiencies, this research utilizes pre-installed CCTVs and proposes a multi-zone people counting methodology. The research process consists of five main processes as follows. First, a single-zone people counting method is developed in order to count people within a singular CCTV. Second, a CCTV code classification system is established to form a horizontal and vertical relationship with the nearby CCTVs, which can be applied in any building. Third, a multi-zone counting algorithm inspired by queues and buffers is developed in order to consider the contextual information from nearby CCTVs and adjust the miscounts from each CCTV. Fourth, equations considering the context of building disaster were formed. Lastly, validation was conducted by performing people counting on an actual non-residential building. The experiment results showed that this multi-zone people counting algorithm can be effective in improving the people counting accuracy in a building, and can be applied in building disaster context.
-
dc.description.abstract건물 재난은 지속적으로 발생해왔으며, 건물 재난 발생 시 제한된 정보를 바탕으로 탐색구조가 이루어져 인명 피해가 발생하고 있다. 따라서 건물 재난이 발생하기 전 평소부터 건물, 층 및 방 별로 인원계수를 하여 재실자 수 정보를 보유한 뒤, 재난이 발생한 시점에 구조대원들에게 그 정보를 전달하는 것이 중요하다.
인원계수 정보를 사전에 보유할 필요성은 피플카운팅 분야에서 연구되어오고 있다. 그러나 이전 연구에서의 인원계수 범위는 건물 전체에 센서를 설치해야 하는 한계와 일반화를 시키기 어려운 점 때문에 단일 또는 소수의 구역으로만 진행되었다. 이러한 한계를 극복하기 위해 본 연구는 기존 건물에 존재하는 CCTV를 활용하여 건물 구역별 인원 계수 방법론을 제안하고자 한다. 연구는 크게 다섯 가지 단계로 구성되어 있다. 첫째, 단일 CCTV 구역에서 사람을 계산하기 위한 모델을 개발한다. 둘째, CCTV 코드 분류체계를 구축하여 인근 CCTV와의 수평 및 수직 관계를 형성하고, 이를 모든 건물에 적용할 수 있도록 한다. 셋째, 대기열과 버퍼의 개념을 활용한 구역별 인원 계수 알고리즘을 개발하여 이를 통해 단일 CCTV에서 발생하는 계수 오차를 줄인다. 넷째, 건물 재난을 위한 인원 계수 방정식을 형성하였다. 마지막으로, 실제 비주거 건물에서 인원 계수를 수행하여 인원 계수 방법론의 실효성을 검증한다. 그 결과 본 연구에서 제안한 방법을 통해 건물 구역별 인원 계수 정확도를 향상시키는데 효과적일 수 있으며, 건물 재난 상황에도 적용될 수 있음을 보여준다.
-
dc.description.tableofcontentsChapter 1 Introduction 1
1.1 Research Background 1
1.2 Problem Statement 5
1.3 Research Objectives 6
1.4 Research Scope 7
1.5 Dissertation Outline 8


Chapter 2 People Counting System 10
2.1 Level of Detail for People Counting 10
2.2 People Counting Sensors 13
2.2.1 Explicit People Counting Sensors 13
2.2.2 Implicit People Counting Sensors 13
2.2.3 Summary of People Counting Sensors 14
2.3 Prior Studies using People Counting Sensors 17


Chapter 3 Single-Zone People Counting 20
3.1 Multi-Zone People Counting Methodology Overview 21
3.2 Single-Zone People Counting Overview 21
3.3 People Detection Module 22
3.3.1 People Detection Module Overview 22
3.3.2 Single-Shot Multibox Detector (SSD) 22
3.3.3 Data Augmentation 25

3.3.4 Parameters in People Detection Module 27
3.4 People Tracking Module 30
3.4.1 People Tracking Module Overview 30
3.4.2 Parameters in People Tracking Module 31
3.5 People Counting Module 33
3.5.1 People Counting Module Overview 33
3.5.2 Example of People Counting Module 33
3.5.3 Parameters in People Tracking Module 34
3.6 Summary of the Modules 37


Chapter 4 Multi-Zone People Counting 38
4.1 CCTV Code Classification 38
4.1.1 Definition of Building Facilities 38
4.1.2 CCTV Code Generation Method 42
4.1.3 Application of CCTV Code Generation Method 44
4.2 Multi-Zone People Counting Algorithm 52
4.2.1 Multi-Zone People Counting Algorithm Overview 52
4.2.2 Utilization of Multi-Zone Context 52
4.2.3 Queue-Buffer Algorithm 55
4.2.4 People Counting Equations 61


Chapter 5 Experimental Validation 62
5.1 Experimental Overview 62
5.2 Single-Zone People Counting 63
5.2.1 Experimental Setup and Data Collection 63
5.2.2 Module Configuration and Parameter Setting 65
5.2.3 Evaluation Metrics 66
5.2.4 Experimental Results and Discussion 67
5.3 Multi-Zone People Counting: Validation 73
5.3.1 Experimental Setup and Data Collection 73
5.3.2 Module Configuration and Parameter Setting 73
5.3.3 Counting Methods and Evaluation Metrics 73
5.3.4 Experimental Results and Discussion 74
5.4 Multi-Zone People Counting: Application 79
5.4.1 Experimental Setup and Data Collection 79
5.4.2 Module Configuration and Parameter Setting 79
5.4.3 Counting Methods and Evaluation Metrics 79
5.4.4 Experimental Results and Discussion 80


Chapter 6 Conclusion 85
6.1 Summary and Contributions 85
6.2 Limitations and Future Study 87


Bibliography 88

Abstract (Korean) 96
-
dc.format.extentv, 97-
dc.language.isoeng-
dc.publisher서울대학교 대학원-
dc.subjectPeople Counting-
dc.subjectSearch and Rescue-
dc.subjectBuilding Disaster-
dc.subjectCCTV-
dc.subjectCode Classification System-
dc.subjectMulti-Zone Counting-
dc.subject인원 계수-
dc.subject탐색구조-
dc.subject건물 재난-
dc.subject코드분류체계-
dc.subject구역별 계수-
dc.subject.ddc624-
dc.titleDeveloping a Multi-Zone People Counting Methodology using Surveillance Cameras for Search and Rescue Efforts during Building Disasters-
dc.title.alternative건물재난 탐색구조를 위한 다중 CCTV 기반 인원 계수 방법론-
dc.typeThesis-
dc.typeDissertation-
dc.contributor.AlternativeAuthorPARK Chan Hyuk-
dc.contributor.department공과대학 건설환경공학부-
dc.description.degreeMaster-
dc.date.awarded2021-02-
dc.identifier.uciI804:11032-000000163809-
dc.identifier.holdings000000000044▲000000000050▲000000163809▲-
Appears in Collections:
Files in This Item:

Altmetrics

Item View & Download Count

  • mendeley

Items in S-Space are protected by copyright, with all rights reserved, unless otherwise indicated.

Share