Smoke and fire detection for intelligent video surveillance using randomness test on multiple features

Smoke and fire detection for intelligent video surveillance using randomness test on multiple features
Issue Date
With the increasing development of surveillance system and the thriving demand of intelligent monitoring, more and more attentions are focused to camera-based emergency detection recently so as to decrease property loss and life loss, since camera-based detection approach can give an earlier warning than other sensor-based approaches, and the result can be confirmed remotely through network with live video. Although many research have been done for fire and smoke detection using camera, the presented approaches are not suitable to be employed in real-environment applications because of the complexity of weather conditions and environment situations. Instead of detecting fire/smoke using heuristic thresholds or fire/smoke patterns found from some specified DB videos, this thesis focuses on the more general and robust fire/smoke features and detects fire/smoke based on the features which are suitable to be adopted generally in real-environment applications. In this study, novel methods
Abstract v ACKNOWLEDGEMENT vii List of Figures x List of Tables xiii Chapter 1 1 INTRODUCTION 1 I. Background 1 II. Purpose and scope 5 Chapter 2 8 Fire Detection 8 I. Related Work 8 I.1 Fire-Colored Pixels Detection 9 I.2 Moving Pixels Detection 11 I.3 Feature Extraction and Classification 13 II. Randomness Test-based Fire Detection 14 II.1. Fire Color Probability Estimation using Gaussian 15 II.2. Motion Probability Estimation for Fire 23 II.3 Robust Features Extraction and Randomness Test 25 II.4 Fire Probability Estimation using Convolution 34 III. Experimental Results 35 IV. Summary 42 Chapter 3 43 Smoke Detection 43 I. Related Work 43 I.1 Motion Detection 44 I.2 Smoke Color Detection 45 I.3 Candidate Smoke Verification 46 II. Probability-based Smoke Detection 47 II.1 Motion Probability Estimation using GMM 48 II.2 Smoke Color Probability Estimation using Increment 51 II.3 Feature Extraction and Verification using SVM 55 II.4 Smoke Probability Estimation using Convolution 67 III. Experimental Results 69 IV. Summary 76 Chapter 4 77 CONCLUSIONS 77 REFERENCES 79
Appears in Collections:
Graduate School (일반대학원) > Theses(로봇공학 석박사 학위논문)
Files in This Item:

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