應用視訊序列之視覺式火焰偵測

Vision-based Fire Detection Using Video Sequences

謝東哲(Tung-Che Hsieh)

指導教授 ： 張元翔

中原大學/電機資訊學院/通訊工程碩士學位學程/碩士(2014年)

https://doi.org/10.6840/cycu201400483

摘要

一直以來火災總是威脅著人們的安全、生命、財產等。大多數火焰偵測傳統設備有其限制性，需要一定程度的火勢才能偵測到。另外這些設備也無法有效運用於戶外空曠地方。近年來，隨著監視錄影設備快速普及，以及影像處理、視訊處理、電腦視覺的發展越來越成熟。將現有的監控攝影設備和基於視覺型的火焰偵測的演算法合併，不只成本相對較低，亦可達到室內室外皆適用，且即早偵測的優點。本研究旨在發展一套「火焰偵測系統」。研究方法包括：(1)影像前置處理(Image Preprocessing)；(2)前景區域分析(Foreground Region Analysis)；(3)火焰動態行為分析(Fire Dynamic Behavior Analysis)；及(4)火焰流動能量分析(Fire Flow Energy Analysis)。研究結果顯示，我們的系統不論是在室內或室外，均可達到至少91%的偵測率。此外，一旦火焰出現，我們的系統能夠在1秒內偵測出火焰(平均延遲約25張影格)。總結而言，本研究提出一套有效且快速的火焰偵測系統，可用於監控系統，以即早偵測火焰避免危害人們的安全、生命、財產等。

關鍵字

視訊分析 ； 影像處理 ； 火焰偵測 ； 渦流效應

並列摘要

Fire, if improperly used, could pose great threats to peoples’ security, life, and property. Motivated by the requirement to detect fire at its early stage, we aimed to develop an automatic system for vision-based fire detection using video sequences. Our system included four major steps, namely image preprocessing, foreground region analysis, fire dynamic behavior analysis, and fire flow energy analysis. Overall, our system could achieve the detection rates of over 91% in either indoor or outdoor environments. In addition, our system could achieve the system response time within 1 second (average delay of ~25 frames) once the fire occurred. In summary, our system could be used in surveillance systems, leading to prevent damage to peoples’ security, life, and property.

並列關鍵字

Fire detection ； Image processing ； Eddies Effect ； Video analysis

參考文獻

[8] T. H. Chen, P.H. Wu, and Y. C. Chiou, “An early fire-detection method based on image processing,” Image Processing 2004, Vol. 3, pp. 1707 - 1710,October 2004.

[1]內政部消防署全球資訊網

http://www.nfa.gov.tw/main/Unit.aspx?ID=&MenuID=378&ListID=129, April 2014.

[2] S. Rinsurongkawong, M. Ekpanyapong, and M. Dailey, “Fire Detection For Early Fire Alarm Based On Optical Flow Video Processing,” IEEE International Conference on Electronic Engineering/Electronics, Computer, Telecommunications and Information Technology(ECTICON), pp. 1 – 4, May 2012.

[3] G. Healey, D. Slater, T. Lin, B. Drda, and D. Goedeke. “A　system for real-time fire detection,” IEEE International Conference on Computer Vision and　Pattern Recognition, pp. 605 – 606, January 1993.