基于视频的烟雾检测算法研究

发布时间：2018-05-01 10:16

本文选题：烟雾检测 + 色调饱和度差分描述子　；参考：《浙江大学》2017年硕士论文

【摘要】：火灾是严重威胁公众生命与财产安全的主要灾害之一,传统的火灾探测技术存在适用范围小、价格高、误报率高、响应速度慢以及提供灾情信息少等弊端。针对传统火灾探测技术存在的缺陷,基于视频的火灾烟雾探测技术应运而生。基于视频的烟雾检测技术比基于视频的火焰检测技术反应时间更早,消防人员能在火势蔓延之前就可以采取应对措施,更有利于对火灾进行有效控制。因此,基于视频的烟雾检测技术具有重要的研究意义和应用价值。本文对基于视频的烟雾检测算法进行详细研究,并在此基础上提出基于烟雾静态特征和动态特征的烟雾检测算法。实验结果证明了本文烟雾检测算法的有效性和鲁棒性。本文的烟雾检测算法主要分为两部分,包括烟雾候选区域提取算法和候选区域烟雾识别算法。本文的主要研究工作如下:(1)研究基于自适应均值背景建模的背景差分算法。(2)研究基于光度不变性颜色空间和光度增益的运动区域提取算法。(3)根据烟雾像素在色度(C)、亮度(I)和饱和度(S)分量上的分布范围,从运动区域中筛选出烟雾候选区域。(4)结合色调饱和度描述子(HS-Descriptor)和梯度直方图(HOG)特征的原理,提出了基于色调饱和度差分描述子(HS-BS-Descriptor)的烟雾检测算法。(5)结合视频中烟雾的光流场直方图(HOF)特征原理,提出了本文的创新点,基于HS-BS-Descriptor+HOF的烟雾检测算法。(6)研究支持向量机(SVM)以及分类器评价指标,用SVM算法训练烟雾识别分类器。
[Abstract]:Fire is one of the main disasters that seriously threaten the safety of public life and property. The traditional fire detection technology has the disadvantages of small scope of application, high price, high false alarm rate, slow response speed and little disaster information. Aiming at the defects of the traditional fire detection technology, the fire smoke detection technology based on video emerges as the times require. The response time of the video based smoke detection technology is earlier than that of the video based flame detection technology. Firefighters can take countermeasures before the fire spreads, which is more conducive to the effective control of the fire. Therefore, video-based smoke detection technology has important research significance and application value. In this paper, the video based smoke detection algorithm is studied in detail, and a smoke detection algorithm based on static and dynamic features of smoke is proposed. Experimental results show the effectiveness and robustness of the proposed smoke detection algorithm. The smoke detection algorithm is mainly divided into two parts, including smoke candidate region extraction algorithm and candidate region smoke recognition algorithm. The main work of this paper is as follows: (1) the background difference algorithm based on adaptive mean value is studied.) the algorithm of extracting moving region based on photometric invariant color space and photometric gain is studied. The range of distributions on the components of C, I and S), The principle of combining HS-Descriptor) with HS-Descriptor and gradient histogram (hog) features is selected from the motion region. A smoke detection algorithm based on HS-BS-Descriptor (HS-BS-Descriptor) is proposed, which combines with the principle of optical flow field histogram (HOF) feature of smoke in video, and the innovation of this paper is presented. The support vector machine (SVM) and the evaluation index of the classifier are studied based on the smoke detection algorithm of HS-BS-Descriptor HOF. The SVM algorithm is used to train the smoke recognition classifier.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【相似文献】