基于机器视觉的机床火灾自动报警技术研究

发布时间：2018-08-24 13:37

【摘要】：随着国家经济的快速发展,时常发生的火灾给工业生产带来了巨大的损失。传统的火灾探测传感器虽然对火灾的预防具有重要的意义,但它们有很多的局限性,尤其在复杂的工厂生产环境中。为了解决传统的检测传感器在一些环境的使用缺陷,国内外很多学者对视频火灾检测算法进行了研究。本文针对火灾检测技术的研究进行了如下工作:(1)首先对火灾检测技术的发展现状进行综述,分析了传统的火灾探测技术的缺陷,然后分析了国内外机器视觉火灾探测发展现状,介绍了各个学者提出的视频图像检测算法,最后得出这些技术只能在某些特殊的场合应用。(2)针对颜色检测算法、移动物体检测算法和红外光谱火焰检测算法进行研究。其中颜色检测算法是基于统计学原理的方法;移动物体检测主要包括差分法、光流法、高斯背景减除算法;本文对其进行了深入的推导。(3)针对火焰的特征分类算法进行研究,在总结这些算法的基础上,我们提出了一种火灾检测方法,该方法结合背景减除算法和区域协方差算子,首先用颜色分布模型和自适应的背景减除算法对视频图像进行预处理,然后提取时空协方差矩阵。最后用支持向量机对视频数据进行分类,得出火焰区域。并和已有文献中的算法效果进行了对比分析。(4)针对机床现在应用要求,开发了一套机器视觉检测系统,详细介绍了机器视觉的机床火灾控制报警系统的系统结构、软件系统、硬件系统。该软件系统已经在北京电加工研究所现场实际运行。经测试,该软件系统算法可靠性强、探测率高。能够满足电加工机床现场需要。本文提出了一种火灾检测方法,该方法结合背景减除算法和区域协方差算子。最后用支持向量机对视频数据进行分类,得出火焰区域。针对北京电加工研究所的特殊应用,开发了一套针对机床火灾的机器视觉系统。
[Abstract]:With the rapid development of national economy, frequent fires have brought huge losses to industrial production. Although the traditional fire detection sensors are of great significance to fire prevention, they have many limitations, especially in the complex production environment of factories. In order to solve the defects of traditional detection sensors in some environments, many scholars at home and abroad have studied the video fire detection algorithm. In this paper, the research work of fire detection technology is as follows: (1) the development of fire detection technology is summarized, the defects of traditional fire detection technology are analyzed, and the present situation of fire detection by machine vision at home and abroad is analyzed. This paper introduces the video image detection algorithms proposed by various scholars, and concludes that these techniques can only be applied in some special situations. (2) the color detection algorithm, moving object detection algorithm and infrared spectrum flame detection algorithm are studied. Color detection algorithm is based on the principle of statistics; moving object detection mainly includes difference method, optical flow method, Gao Si background subtraction algorithm. On the basis of summarizing these algorithms, we propose a fire detection method, which combines background subtraction algorithm and regional covariance operator. Firstly, the color distribution model and adaptive background subtraction algorithm are used to preprocess the video image. Then the space-time covariance matrix is extracted. Finally, the support vector machine is used to classify the video data and the flame region is obtained. And compared with the existing literature algorithm results. (4) according to the current application requirements of machine tools, a set of machine vision detection system is developed. The system structure and software system of machine tool fire control and alarm system based on machine vision are introduced in detail. Hardware system. The software system has been in practical operation in Beijing Institute of Electrical processing. After testing, the software system has strong reliability and high detectability. Able to meet the field needs of electrical machining machine tools. In this paper, a fire detection method is proposed, which combines background subtraction algorithm and regional covariance operator. Finally, the support vector machine is used to classify the video data and the flame region is obtained. Aiming at the special application of Beijing Institute of Electrical processing, a machine vision system for machine tool fire is developed.
【学位授予单位】：沈阳理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TG502.39;TP391.41

【参考文献】