【摘要】：火災(zāi)這一自然災(zāi)害對于人們?nèi)粘５纳a(chǎn)生活來說具有非常大的破壞力,同時也對人民群眾的財產(chǎn)安全和生命安全造成了嚴重的威脅。通�；馂�(zāi)發(fā)生的初期破壞力不大,容易被忽視,但其發(fā)展迅速,隨時間推移,破壞力迅速增長,因此越早發(fā)現(xiàn)火情并將其消除,所遭受的損失就越小。傳統(tǒng)的火情監(jiān)測方法一般依賴于人工巡檢,其成本巨大,且在某些特殊危險區(qū)域無法使用;而另外一種更廣泛的火情監(jiān)測方法是基于多種傳感器技術(shù):將感煙、感光、感氣等傳感器與網(wǎng)絡(luò)技術(shù)結(jié)合,通過對監(jiān)測場所的物理和化學物質(zhì)進行遠程監(jiān)測分析來實現(xiàn)火情監(jiān)測。這種方法雖然克服了人工巡檢的缺點,但是容易受環(huán)境影響,難以勝任開闊場所的監(jiān)測任務(wù)。近年來,隨著數(shù)字圖像處理技術(shù)以及計算機視覺研究領(lǐng)域的不斷發(fā)展,融合了數(shù)字圖像處理和計算機視覺技術(shù)的基于視頻的火情監(jiān)測技術(shù)由于成本低廉、適用范圍廣等優(yōu)點吸引了國內(nèi)外大量的專家學者,成為研究熱點。本文主要研究分析了傳統(tǒng)火情監(jiān)測技術(shù)的研究現(xiàn)狀以及不足之處,并在對視頻序列中的火焰圖像在空間和時間上的特征進行分析后,提出了運用數(shù)字圖像處理和計算機視覺處理技術(shù)的基于視頻特征的火情監(jiān)測方法。具體內(nèi)容如下:首先針對視頻中運動火焰前景提取的需求提出一種改進的ViBe(visual background extractor)算法。將亮度特征匹配引入背景模型更新策略,以此來解決原始ViBe算法無法區(qū)分運動火焰和其他運動物體,且在光線變化時容易誤檢測的問題;同時,利用幀間差分法計算速度快的特點預(yù)提取出畫面中的運動區(qū)域以及背景作為初始化幀,再使用ViBe算法對該區(qū)域做進一步更準確的前景提取,改善了ViBe算法實時性隨分辨率增大而下降,且初始化幀選取不當時易出現(xiàn)檢測鬼影的不足。其次對提取出的前景區(qū)域進行火焰顏色特征匹配和火焰形態(tài)特征匹配。提出了基于正序數(shù)比值的火焰尖角程度匹配模型;并在分析火焰區(qū)域面積增長性的特征后設(shè)計了一種基于面積增長閾值的檢測報警算法。最后,論文基于Qt和OpenCv開發(fā)了火情監(jiān)測系統(tǒng)軟件,對提出的相關(guān)算法進行了驗證。實驗表明,本系統(tǒng)能夠較好的識別出視頻中的火焰信息,具有良好的工程應(yīng)用前景。
文內(nèi)圖片：
圖片說明：中值濾波常用窗口Fig2-2Commonwindowsinmediumfilter
[Abstract]:Fire, a natural disaster, is very destructive to people's daily production and life, and also poses a serious threat to the property safety and life safety of the people. Usually, the initial destructive power of fire is not great and easy to be ignored, but its development is rapid. With the passage of time, the destructive power increases rapidly, so the sooner the fire is discovered and eliminated, the smaller the loss will be. The traditional fire monitoring method generally depends on manual inspection, which is costly and can not be used in some special dangerous areas. Another more extensive fire monitoring method is based on a variety of sensor technologies: combining smoke sensing, photosensory, gas sensing and other sensors with network technology to achieve fire monitoring by remote monitoring and analysis of physical and chemical substances in the monitoring place. Although this method can overcome the shortcomings of manual inspection, it is easy to be affected by the environment and is difficult to be competent for the monitoring task of open places. In recent years, with the continuous development of digital image processing technology and computer vision research field, video-based fire monitoring technology, which combines digital image processing and computer vision technology, has attracted a large number of experts and scholars at home and abroad because of its low cost and wide application range. In this paper, the research status and shortcomings of traditional fire monitoring technology are studied and analyzed. After analyzing the spatial and temporal characteristics of flame images in video sequences, a fire monitoring method based on video features using digital image processing and computer vision processing technology is proposed. The main contents are as follows: firstly, an improved ViBe (visual background extractor) algorithm is proposed to meet the requirements of moving flame foreground extraction in video. The brightness feature matching is introduced into the background model updating strategy to solve the problem that the original ViBe algorithm can not distinguish moving flame from other moving objects, and it is easy to misdetect when the light changes. At the same time, the moving region and background in the picture are taken as the initial frame by using the inter-frame difference method to calculate the fast speed, and then the ViBe algorithm is used to extract the foreground more accurately, which improves the real-time performance of ViBe algorithm decreases with the increase of resolution, and the lack of ghost detection is easy to occur when the initial frame is not selected. Secondly, flame color feature matching and flame morphological feature matching are carried out for the extracted foreground region. A flame angle matching model based on positive ordinal ratio is proposed, and a detection and alarm algorithm based on area growth threshold is designed after analyzing the characteristics of flame area growth. Finally, the fire monitoring system software is developed based on Qt and OpenCv, and the related algorithms are verified. The experimental results show that the system can better identify the flame information in the video and has a good engineering application prospect.
【學位授予單位】：江西農(nóng)業(yè)大學
【學位級別】：碩士
【學位授予年份】：2017
【分類號】：X932;TP391.41

【參考文獻】

相關(guān)期刊論文 前10條

1 吳茜茵;嚴云洋;杜靜;高尚兵;劉以安;;多特征融合的火焰檢測算法[J];智能系統(tǒng)學報;2015年02期

2 張生杰;張為;;基于混合高斯模型和幀差的火焰前景提取算法[J];信息技術(shù);2014年10期

3 陳星明;廖娟;李勃;陳啟美;;動態(tài)背景下基于改進視覺背景提取的前景檢測[J];光學精密工程;2014年09期

4 高偉偉;曹江濤;;一種基于RGB與HIS顏色空間模型的火焰尺寸檢測方法研究[J];遼寧石油化工大學學報;2014年04期

5 耿慶田;于繁華;趙宏偉;王闖;;基于顏色特征的火焰檢測新算法[J];吉林大學學報(工學版);2014年06期

6 嚴紅亮;王福龍;劉志煌;沈士忠;;基于ViBe算法的改進背景減去法[J];計算機系統(tǒng)應(yīng)用;2014年06期

7 胡小冉;孫涵;;一種新的基于ViBe的運動目標檢測方法[J];計算機科學;2014年02期

8 陳亮;陳曉竹;范振濤;;基于Vibe的鬼影抑制算法[J];中國計量學院學報;2013年04期

9 吳桐;王玲;;基于幀差分塊的混合高斯背景模型[J];計算機工程與應(yīng)用;2014年23期

10 袁國武;陳志強;龔健;徐丹;廖仁健;何俊遠;;一種結(jié)合光流法與三幀差分法的運動目標檢測算法[J];小型微型計算機系統(tǒng);2013年03期

，

本文編號：2510253