基于视频特征的火情监测研究
发布时间：2019-07-04 21:42
【摘要】：火灾这一自然灾害对于人们日常的生产生活来说具有非常大的破坏力,同时也对人民群众的财产安全和生命安全造成了严重的威胁。通常火灾发生的初期破坏力不大,容易被忽视,但其发展迅速,随时间推移,破坏力迅速增长,因此越早发现火情并将其消除,所遭受的损失就越小。传统的火情监测方法一般依赖于人工巡检,其成本巨大,且在某些特殊危险区域无法使用;而另外一种更广泛的火情监测方法是基于多种传感器技术:将感烟、感光、感气等传感器与网络技术结合,通过对监测场所的物理和化学物质进行远程监测分析来实现火情监测。这种方法虽然克服了人工巡检的缺点,但是容易受环境影响,难以胜任开阔场所的监测任务。近年来,随着数字图像处理技术以及计算机视觉研究领域的不断发展,融合了数字图像处理和计算机视觉技术的基于视频的火情监测技术由于成本低廉、适用范围广等优点吸引了国内外大量的专家学者,成为研究热点。本文主要研究分析了传统火情监测技术的研究现状以及不足之处,并在对视频序列中的火焰图像在空间和时间上的特征进行分析后,提出了运用数字图像处理和计算机视觉处理技术的基于视频特征的火情监测方法。具体内容如下:首先针对视频中运动火焰前景提取的需求提出一种改进的ViBe(visual background extractor)算法。将亮度特征匹配引入背景模型更新策略,以此来解决原始ViBe算法无法区分运动火焰和其他运动物体,且在光线变化时容易误检测的问题;同时,利用帧间差分法计算速度快的特点预提取出画面中的运动区域以及背景作为初始化帧,再使用ViBe算法对该区域做进一步更准确的前景提取,改善了ViBe算法实时性随分辨率增大而下降,且初始化帧选取不当时易出现检测鬼影的不足。其次对提取出的前景区域进行火焰颜色特征匹配和火焰形态特征匹配。提出了基于正序数比值的火焰尖角程度匹配模型;并在分析火焰区域面积增长性的特征后设计了一种基于面积增长阈值的检测报警算法。最后,论文基于Qt和OpenCv开发了火情监测系统软件,对提出的相关算法进行了验证。实验表明,本系统能够较好的识别出视频中的火焰信息,具有良好的工程应用前景。
文内图片：
图片说明：中值滤波常用窗口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.
【学位授予单位】：江西农业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：X932;TP391.41
本文编号：2510253
文内图片：
图片说明：中值滤波常用窗口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.
【学位授予单位】：江西农业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：X932;TP391.41
【参考文献】
相关期刊论文 前10条
1 吴茜茵;严云洋;杜静;高尚兵;刘以安;;多特征融合的火焰检测算法[J];智能系统学报;2015年02期
2 张生杰;张为;;基于混合高斯模型和帧差的火焰前景提取算法[J];信息技术;2014年10期
3 陈星明;廖娟;李勃;陈启美;;动态背景下基于改进视觉背景提取的前景检测[J];光学精密工程;2014年09期
4 高伟伟;曹江涛;;一种基于RGB与HIS颜色空间模型的火焰尺寸检测方法研究[J];辽宁石油化工大学学报;2014年04期
5 耿庆田;于繁华;赵宏伟;王闯;;基于颜色特征的火焰检测新算法[J];吉林大学学报(工学版);2014年06期
6 严红亮;王福龙;刘志煌;沈士忠;;基于ViBe算法的改进背景减去法[J];计算机系统应用;2014年06期
7 胡小冉;孙涵;;一种新的基于ViBe的运动目标检测方法[J];计算机科学;2014年02期
8 陈亮;陈晓竹;范振涛;;基于Vibe的鬼影抑制算法[J];中国计量学院学报;2013年04期
9 吴桐;王玲;;基于帧差分块的混合高斯背景模型[J];计算机工程与应用;2014年23期
10 袁国武;陈志强;龚健;徐丹;廖仁健;何俊远;;一种结合光流法与三帧差分法的运动目标检测算法[J];小型微型计算机系统;2013年03期
，本文编号：2510253
本文链接：https://www.wllwen.com/kejilunwen/anquangongcheng/2510253.html
