使用隨機森林演算法做煙火偵測

Fire and smoke detection using random forest algorithm

林伯勳(Pa-Hsun Lin)

指導教授 ： 石勝文

國立暨南國際大學/科技學院/資訊工程學系/碩士(2013年)

https://doi.org/10.6837/NCNU.2013.00224

摘要

隨著電腦能力的進展，影像處理的技術和知識更加成熟，以及智能視頻監控式系統的功能也大大擴展。在這篇論文中，我們以隨機森林演算法為基礎，發展出一套基於視頻的
火災和煙霧偵測系統。我們利用煙火明顯的顏色和變化特性來選擇出候選區域，接著，針對候選區域分析紋理和運動模式的特徵來決定煙火區域。我們提出了使用局部二值模式 (LBP) 方法來擷取煙火紋理和運動模式的特徵，利用LBP特徵所訓練出來的隨機森林來做煙火偵測，以減少誤報也加強煙火辨識率。

關鍵字

隨機森林演算法 ； 局部二值模式 ； 煙霧偵測 ； 火災偵測

並列摘要

Along with the progress of computer computation capabilities, sophisticated image processing/understanding methods have been developed and the functions of intelligent video surveillance systems have been greatly extended. In this thesis, we develop a video-based fire and smoke detection system based on the random forest algorithm. We use the distinct color and image variation properties of fire/smoke to select candidate regions. Then, image features of texture and motion patterns of the candidate regions are analyzed to determine any fire/smoke region. We propose to extract the features of both the texture and motion patterns of the fire/smoke with the local binary pattern (LBP) method. The random forest method is augmented to use the LBP features for fire/smoke detection to reduce false positive and enhance the fire and smoke detection rate.

並列關鍵字

Random Forests ； Local Binary Pattern ； Smoke Detection ； Fire Detection

參考文獻

[1] B. U. Toreyin, Y. Dedeo glu, and A. E. Cetin, “Flame detection in video using hidden

markov models,” in Proceedings of the IEEE International Conference on Image Processing,

pp. 1230 – 1233, 2005.

[2] H. Li, Q. Liu, and S. Wang, “A novel fire recognition algorithm based on flame’s multifeatures

fusion,” in Proceedings of the International Conference on Computer Communication